[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"portal-settings:stajic:es":3,"public-menus:all":38,"post:mcp-vs-a2a-vs-ucp-vs-ap2-vs-a2ui-the-agent-protocol-stack-explained:es":205,"related:post:mcp-vs-a2a-vs-ucp-vs-ap2-vs-a2ui-the-agent-protocol-stack-explained:es:1":1826},{"statusCode":4,"data":5,"message":37},200,{"tenantId":6,"lang":7,"defaultLang":8,"siteUrl":9,"contactEmail":10,"brandName":11,"logoUrl":12,"siteName":11,"siteDescription":13,"ogImage":10,"robotsIndex":14,"socialLinks":10,"reservedSlugs":10,"seoPolicy":15},"stajic","es","de","https:\u002F\u002Fstajic.de",null,"Stajic Platform","\u002FLogo_Planet.svg","Stajic Portal",true,{"branding":16,"relatedContent":17,"crossDomainLinks":18},{"logoUrl":12},{"enabled":14},[19,22,25,28,31,34],{"url":20,"label":21,"isActive":14,"showInFooter":14,"includeInSameAs":14},"https:\u002F\u002Ffigure.rocks","figure.rocks",{"url":23,"label":24,"isActive":14,"showInFooter":14,"includeInSameAs":14},"https:\u002F\u002Floving.rocks","loving.rocks",{"url":26,"label":27,"isActive":14,"showInFooter":14,"includeInSameAs":14},"https:\u002F\u002Fbazify.com","bazify.com",{"url":29,"label":30,"isActive":14,"showInFooter":14,"includeInSameAs":14},"https:\u002F\u002Fbazify.de","bazify.de",{"url":32,"label":33,"isActive":14,"showInFooter":14,"includeInSameAs":14},"https:\u002F\u002Fbazify.at","bazify.at",{"url":35,"label":36,"isActive":14,"showInFooter":14,"includeInSameAs":14},"https:\u002F\u002Fbazify.ba","bazify.ba","Portal settings resolved",[39,45],{"id":40,"name":41,"location":42,"isActive":14,"isDefault":43,"items":44},1,"main-navigation","header",false,[],{"id":46,"name":47,"location":48,"isActive":14,"isDefault":14,"items":49},4,"main-menu","sidebar",[50,66,79,93,103,118,133],{"id":51,"title":52,"url":60,"target":61,"icon":62,"isActive":14,"type":63,"productId":10,"categoryId":10,"shopCategoryId":10,"articleId":10,"pageId":64,"portfolioId":10,"children":65},"item-18",{"de":53,"en":54,"es":55,"fr":56,"it":54,"ru":57,"sr":58,"zh":59},"Startseite","Home","Inicio","Accueil","Главная","Почетна","首页","\u002Ffull-stack-web-developer-munich-performance-seo-and-maintainable-builds","_self","i-lucide-home","page",111,[],{"id":67,"title":68,"url":75,"target":61,"icon":76,"isActive":14,"type":63,"productId":10,"categoryId":10,"shopCategoryId":10,"articleId":10,"pageId":77,"portfolioId":10,"children":78},"item-22",{"de":69,"en":69,"es":70,"fr":69,"it":71,"ru":72,"sr":73,"zh":74},"Vision","Visión","Visione","Видение","Визија","想象","\u002Fueber-uns-webdesign-muenchen-webaplikation","i-lucide-eye",113,[],{"id":80,"title":81,"url":89,"target":61,"icon":90,"isActive":14,"type":63,"productId":10,"categoryId":10,"shopCategoryId":10,"articleId":10,"pageId":91,"portfolioId":10,"children":92},"item-19",{"de":82,"en":83,"es":84,"fr":83,"it":85,"ru":86,"sr":87,"zh":88},"Leistungen","Services","Servicios","Servizi","Услуги","Услуге","服务","\u002Fservices-dienstleistungen-muenchen","i-lucide-wrench",116,[],{"id":94,"title":95,"url":99,"target":61,"icon":100,"isActive":14,"type":63,"productId":10,"categoryId":10,"shopCategoryId":10,"articleId":10,"pageId":101,"portfolioId":10,"children":102},"item-23",{"de":96,"en":96,"es":96,"fr":96,"it":96,"ru":97,"sr":97,"zh":98},"Blog","Блог","博客","\u002Fblog","i-lucide-book-open",112,[],{"id":104,"title":105,"url":114,"target":61,"icon":115,"isActive":14,"type":63,"productId":10,"categoryId":10,"shopCategoryId":10,"articleId":10,"pageId":116,"portfolioId":10,"children":117},"item-32",{"de":106,"en":107,"es":108,"fr":109,"it":110,"ru":111,"sr":112,"zh":113},"Neue Technologien","New Technologies","Nuevas tecnologías","Nouvelles technologies","Nuove tecnologie","Новые технологии","Нове технологије","新技术！","\u002Fneue-webtechnologien","i-lucide-sparkles",122,[],{"id":119,"title":120,"url":129,"target":61,"icon":130,"isActive":14,"type":63,"productId":10,"categoryId":10,"shopCategoryId":10,"articleId":10,"pageId":131,"portfolioId":10,"children":132},"item-20",{"de":121,"en":122,"es":123,"fr":124,"it":125,"ru":126,"sr":127,"zh":128},"Kontakt","Contact us!","Contacto","Contact","Contatto","Контакт","Контактирајте нас","联系我们！","\u002Fcontact","i-lucide-mail",115,[],{"id":134,"title":135,"url":144,"target":61,"icon":145,"isActive":14,"type":63,"productId":10,"categoryId":10,"shopCategoryId":10,"articleId":10,"pageId":146,"portfolioId":10,"children":147},"item-21",{"de":136,"en":137,"es":138,"fr":139,"it":140,"ru":141,"sr":142,"zh":143},"Unsere Arbeit","Our Work","Nuestro trabajo","Nos réalisations","I nostri lavori","Наши работы","Наши радови","文件夹","\u002Fportfolio","i-lucide-briefcase",114,[148,161,175,181,193],{"id":149,"title":150,"url":144,"target":61,"icon":159,"isActive":14,"type":63,"productId":10,"categoryId":10,"shopCategoryId":10,"articleId":10,"pageId":146,"portfolioId":10,"children":160},"item-24",{"de":151,"en":152,"es":153,"fr":154,"it":155,"ru":156,"sr":157,"zh":158},"Alle Projekte","All Projects","Todos los proyectos","Tous les projets","Tutti i progetti","Все проекты","Сви пројекти","所有项目","i-lucide-grid-3x3",[],{"id":162,"title":163,"url":171,"target":61,"icon":172,"isActive":14,"type":173,"productId":10,"categoryId":10,"shopCategoryId":10,"articleId":10,"pageId":10,"portfolioId":10,"children":174},"item-29",{"de":164,"en":165,"es":166,"fr":167,"it":168,"ru":169,"sr":170,"zh":143},"Local Roots, Global Reach","Local Roots - Global Reach","Empresa local ","Entreprise locale","Azienda locale","Местная компания","Локално предузеће глобално тржиште","\u002Fportfolio\u002Flocal-roots-global-reach-communication-media-systems-for-modern-business","i-lucide-folder","custom",[],{"id":176,"title":177,"url":179,"target":61,"icon":172,"isActive":14,"type":173,"productId":10,"categoryId":10,"shopCategoryId":10,"articleId":10,"pageId":10,"portfolioId":10,"children":180},"item-28",{"de":178,"en":178,"es":178,"fr":178,"it":178,"ru":178,"sr":178,"zh":178},"Solr Suggester","\u002Fportfolio\u002Fsolr-fuzzy-suggester-und-solr-infix-suggester-abfrage-ueber-ajax-und-filterung",[],{"id":182,"title":183,"url":191,"target":61,"icon":172,"isActive":14,"type":173,"productId":10,"categoryId":10,"shopCategoryId":10,"articleId":10,"pageId":10,"portfolioId":10,"children":192},"item-27",{"de":184,"en":185,"es":186,"fr":187,"it":188,"ru":189,"sr":190,"zh":185},"Firmenwebseite SEO","Company Website SEO","Sitio web corporativo SEO","Site web d’entreprise SEO","Sito web aziendale SEO","Корпоративный сайт SEO","Пословна веб-страница SEO","\u002Fportfolio\u002Fseo-sem-branding-mobile-webseite-muenchen",[],{"id":194,"title":195,"url":203,"target":61,"icon":172,"isActive":14,"type":173,"productId":10,"categoryId":10,"shopCategoryId":10,"articleId":10,"pageId":10,"portfolioId":10,"children":204},"item-31",{"de":196,"en":197,"es":198,"fr":199,"it":200,"ru":201,"sr":202,"zh":197},"Digitalisierungsportal","Digitalization Portal","Portal de digitalización","Portail de numérisation","Portale di digitalizzazione","Портал цифровизации","Портал за дигитализацију","\u002Fportfolio\u002Fdigitalisierungsportal-archiv-museum-bibliothek-ead-lido-mets-mods",[],{"statusCode":4,"data":206,"message":1825},{"id":207,"title":208,"slug":209,"content":210,"contentJson":211,"excerpt":896,"featuredImage":897,"featuredImageAlt":898,"featuredImageCaption":10,"featuredImageTitle":10,"featuredImageCopyright":10,"featuredImageAuthor":10,"featuredImageSourceUrl":10,"featuredImageLicense":10,"featuredImageIsAiGenerated":43,"status":899,"publishedAt":900,"createdAt":901,"updatedAt":902,"seoLocalePaths":903,"categories":912,"author":925,"translations":930},"476","MCP vs A2A vs UCP vs AP2 vs A2UI: La pila de protocolos de agentes explicada","mcp-vs-a2a-vs-ucp-vs-ap2-vs-a2ui-the-agent-protocol-stack-explained","\u003Cnav class=\"editorjs-toc\" data-editorjs-toc=\"true\" aria-label=\"Contenido\">\u003Cstrong class=\"editorjs-toc__title\">Contenido\u003C\u002Fstrong>\u003Col class=\"editorjs-toc__list editorjs-toc__list--depth-0\">\u003Cli class=\"editorjs-toc__item\">\u003Ca href=\"#section-6\" class=\"editorjs-toc__link\">El error fundamental: comparar protocolos que operan en límites diferentes\u003C\u002Fa>\u003C\u002Fli>\u003Cli class=\"editorjs-toc__item\">\u003Ca href=\"#section-9\" class=\"editorjs-toc__link\">La Pila de Responsabilidad de Protocolos\u003C\u002Fa>\u003C\u002Fli>\u003Cli class=\"editorjs-toc__item\">\u003Ca href=\"#section-12\" class=\"editorjs-toc__link\">1. MCP: conectar el agente a las capacidades\u003C\u002Fa>\u003Col class=\"editorjs-toc__list editorjs-toc__list--depth-1\">\u003Cli class=\"editorjs-toc__item\">\u003Ca href=\"#section-15\" class=\"editorjs-toc__link\">Utilice MCP cuando\u003C\u002Fa>\u003C\u002Fli>\u003C\u002Fol>\u003C\u002Fli>\u003Cli class=\"editorjs-toc__item\">\u003Ca href=\"#section-18\" class=\"editorjs-toc__link\">2. A2A: conectar agentes independientes\u003C\u002Fa>\u003Col class=\"editorjs-toc__list editorjs-toc__list--depth-1\">\u003Cli class=\"editorjs-toc__item\">\u003Ca href=\"#section-22\" class=\"editorjs-toc__link\">Use A2A cuando\u003C\u002Fa>\u003C\u002Fli>\u003C\u002Fol>\u003C\u002Fli>\u003Cli class=\"editorjs-toc__item\">\u003Ca href=\"#section-24\" class=\"editorjs-toc__link\">MCP vs. A2A: integración vertical frente a colaboración horizontal\u003C\u002Fa>\u003C\u002Fli>\u003Cli class=\"editorjs-toc__item\">\u003Ca href=\"#section-27\" class=\"editorjs-toc__link\">3. UCP: estandarizar el comercio agéntico\u003C\u002Fa>\u003Col class=\"editorjs-toc__list editorjs-toc__list--depth-1\">\u003Cli class=\"editorjs-toc__item\">\u003Ca href=\"#section-31\" class=\"editorjs-toc__link\">Use UCP cuando\u003C\u002Fa>\u003C\u002Fli>\u003C\u002Fol>\u003C\u002Fli>\u003Cli class=\"editorjs-toc__item\">\u003Ca href=\"#section-33\" class=\"editorjs-toc__link\">4. AP2: demostrar que el agente tenía autorización para gastar\u003C\u002Fa>\u003C\u002Fli>\u003Cli class=\"editorjs-toc__item\">\u003Ca href=\"#section-37\" class=\"editorjs-toc__link\">UCP vs. AP2: semántica de transacciones frente a autoridad\u003C\u002Fa>\u003C\u002Fli>\u003Cli class=\"editorjs-toc__item\">\u003Ca href=\"#section-39\" class=\"editorjs-toc__link\">5. A2UI: permita que los agentes describan interfaces sin adueñarse de su frontend\u003C\u002Fa>\u003Col class=\"editorjs-toc__list editorjs-toc__list--depth-1\">\u003Cli class=\"editorjs-toc__item\">\u003Ca href=\"#section-43\" class=\"editorjs-toc__link\">Utilice A2UI cuando\u003C\u002Fa>\u003C\u002Fli>\u003C\u002Fol>\u003C\u002Fli>\u003Cli class=\"editorjs-toc__item\">\u003Ca href=\"#section-45\" class=\"editorjs-toc__link\">La prueba de selección de protocolos\u003C\u002Fa>\u003C\u002Fli>\u003Cli class=\"editorjs-toc__item\">\u003Ca href=\"#section-48\" class=\"editorjs-toc__link\">Un flujo de trabajo multiprotocolo realista\u003C\u002Fa>\u003C\u002Fli>\u003Cli class=\"editorjs-toc__item\">\u003Ca href=\"#section-50\" class=\"editorjs-toc__link\">Por qué es poco probable que un protocolo universal de agentes los reemplace a todos\u003C\u002Fa>\u003C\u002Fli>\u003Cli class=\"editorjs-toc__item\">\u003Ca href=\"#section-54\" class=\"editorjs-toc__link\">La composición de protocolos genera nuevos modos de fallo\u003C\u002Fa>\u003C\u002Fli>\u003Cli class=\"editorjs-toc__item\">\u003Ca href=\"#section-56\" class=\"editorjs-toc__link\">La elección del protocolo no reemplaza la arquitectura de la aplicación\u003C\u002Fa>\u003C\u002Fli>\u003Cli class=\"editorjs-toc__item\">\u003Ca href=\"#section-60\" class=\"editorjs-toc__link\">¿Qué cambiaría esta respuesta?\u003C\u002Fa>\u003C\u002Fli>\u003Cli class=\"editorjs-toc__item\">\u003Ca href=\"#section-63\" class=\"editorjs-toc__link\">Limitaciones\u003C\u002Fa>\u003C\u002Fli>\u003Cli class=\"editorjs-toc__item\">\u003Ca href=\"#section-66\" class=\"editorjs-toc__link\">Conclusión\u003C\u002Fa>\u003C\u002Fli>\u003Cli class=\"editorjs-toc__item\">\u003Ca href=\"#section-69\" class=\"editorjs-toc__link\">Preguntas frecuentes\u003C\u002Fa>\u003C\u002Fli>\u003Cli class=\"editorjs-toc__item\">\u003Ca href=\"#section-71\" class=\"editorjs-toc__link\">Glosario\u003C\u002Fa>\u003C\u002Fli>\u003Cli class=\"editorjs-toc__item\">\u003Ca href=\"#section-73\" class=\"editorjs-toc__link\">Fuentes principales y lecturas adicionales\u003C\u002Fa>\u003C\u002Fli>\u003C\u002Fol>\u003C\u002Fnav>\n\u003Cp>Los protocolos de agentes de IA se multiplican con rapidez: MCP, A2A, UCP, AP2, A2UI y estándares adyacentes aparecen cada vez más en los mismos diagramas de arquitectura. A menudo se describen como protocolos competidores. En la práctica, la mayoría resuelve diferentes problemas de interoperabilidad en límites distintos. La pregunta útil no es “¿Qué protocolo ganará?”, sino “¿Qué relación en el sistema necesita estandarizarse?”\u003C\u002Fp>\n\u003Caside class=\"editorjs-callout editorjs-callout--info my-6 rounded-xl border p-5 border-blue-300 bg-blue-50 dark:border-blue-900 dark:bg-blue-950\u002F20\" role=\"note\">\u003Cstrong class=\"block mb-2 text-gray-900 dark:text-gray-100\">Respuesta directa\u003C\u002Fstrong>\u003Cdiv class=\"text-gray-700 dark:text-gray-200\"><strong>MCP, A2A, UCP, AP2 y A2UI son en su mayoría complementarios, no sustitutos.<\u002Fstrong> MCP conecta una aplicación de IA con herramientas, datos y recursos. A2A conecta agentes independientes entre sí. UCP estandariza las interacciones de comercio entre interfaces de usuario para consumidores, empresas y ecosistemas de pago. AP2 añade autorización verificable e intención de pago a las transacciones lideradas por agentes. A2UI permite que un agente describa interfaces de usuario interactivas sin enviar código de aplicación arbitrario. Un sistema de producción puede utilizar legítimamente varios de ellos en un mismo flujo de trabajo.\u003C\u002Fdiv>\u003C\u002Faside>\n\u003Caside class=\"editorjs-callout editorjs-callout--warning my-6 rounded-xl border p-5 border-amber-300 bg-amber-50 dark:border-amber-900 dark:bg-amber-950\u002F20\" role=\"note\">\u003Cstrong class=\"block mb-2 text-gray-900 dark:text-gray-100\">Estándares en rápida evolución\u003C\u002Fstrong>\u003Cdiv class=\"text-gray-700 dark:text-gray-200\">Este artículo refleja el panorama de protocolos a fecha de <strong>25 de septiembre de 2026<\u002Fstrong>. A2A ha alcanzado la versión 1.0, la línea actual de MCP TypeScript v2 implementa la especificación del 2026-07-28, UCP ya ha añadido versiones de protocolo más recientes de 2026 y A2UI continúa evolucionando. Siempre verifique la especificación actual antes de la implementación.\u003C\u002Fdiv>\u003C\u002Faside>\n\u003Caside class=\"editorjs-callout editorjs-callout--note my-6 rounded-xl border p-5 border-gray-300 bg-gray-50 dark:border-gray-700 dark:bg-gray-900\u002F40\" role=\"note\">\u003Cstrong class=\"block mb-2 text-gray-900 dark:text-gray-100\">El modelo utilizado en este artículo\u003C\u002Fstrong>\u003Cdiv class=\"text-gray-700 dark:text-gray-200\">La Pila de Responsabilidad de Protocolos y la Prueba de Selección de Protocolos a continuación son modelos prácticos de arquitectura propuestos aquí. No constituyen terminología oficial de los proyectos de los protocolos.\u003C\u002Fdiv>\u003C\u002Faside>\n\u003Ch2 id=\"section-6\">El error fundamental: comparar protocolos que operan en límites diferentes\u003C\u002Fh2>\n\u003Cp>Un protocolo es útil porque dos sistemas implementados de forma independiente necesitan un contrato estable. El contrato solo tiene sentido si el límite está claro. Un agente que se comunica con una base de datos tiene un problema de interoperabilidad diferente al de un agente que delega trabajo en otro, un comprador que autoriza una compra o un agente remoto que solicita a una aplicación nativa que renderice un formulario.\u003C\u002Fp>\n\u003Cp>La guía de desarrolladores de Google de 2026 presenta explícitamente MCP, A2A, UCP, AP2, A2UI y los protocolos de interfaz de usuario relacionados como una pila de estándares complementarios. Un mismo flujo de trabajo de ejemplo puede utilizar varios de ellos de forma conjunta: herramientas para el inventario, agentes remotos para proveedores, comercio para pedidos, autorización de pagos para gastos y protocolos de interfaz de usuario para la interacción.\u003C\u002Fp>\n\u003Ch2 id=\"section-9\">La Pila de Responsabilidad de Protocolos\u003C\u002Fh2>\n\u003Cdiv class=\"overflow-x-auto\">\u003Ctable class=\"w-full border-collapse\">\u003Cthead>\u003Ctr>\u003Cth class=\"border border-gray-300 px-4 py-2 text-left font-semibold\">Protocolo\u003C\u002Fth>\u003Cth class=\"border border-gray-300 px-4 py-2 text-left font-semibold\">¿Qué relación estandariza?\u003C\u002Fth>\u003Cth class=\"border border-gray-300 px-4 py-2 text-left font-semibold\">Abstracción principal\u003C\u002Fth>\u003Cth class=\"border border-gray-300 px-4 py-2 text-left font-semibold\">No diseñado principalmente para\u003C\u002Fth>\u003C\u002Ftr>\u003C\u002Fthead>\u003Ctbody>\u003Ctr>\u003Ctd class=\"border border-gray-300 px-4 py-2\">MCP\u003C\u002Ftd>\u003Ctd class=\"border border-gray-300 px-4 py-2\">Aplicación de IA ↔ herramientas, recursos y datos\u003C\u002Ftd>\u003Ctd class=\"border border-gray-300 px-4 py-2\">Herramientas, recursos, prompts e intercambio de capacidades entre host y servidor\u003C\u002Ftd>\u003Ctd class=\"border border-gray-300 px-4 py-2\">Colaboración independiente entre agentes o semántica comercial\u003C\u002Ftd>\u003C\u002Ftr>\u003Ctr>\u003Ctd class=\"border border-gray-300 px-4 py-2\">A2A\u003C\u002Ftd>\u003Ctd class=\"border border-gray-300 px-4 py-2\">Agente ↔ agente independiente\u003C\u002Ftd>\u003Ctd class=\"border border-gray-300 px-4 py-2\">Descubrimiento de agentes, mensajes, tareas, artefactos y colaboración de larga duración\u003C\u002Ftd>\u003Ctd class=\"border border-gray-300 px-4 py-2\">Integración directa de bases de datos\u002Fherramientas\u003C\u002Ftd>\u003C\u002Ftr>\u003Ctr>\u003Ctd class=\"border border-gray-300 px-4 py-2\">UCP\u003C\u002Ftd>\u003Ctd class=\"border border-gray-300 px-4 py-2\">Superficie de consumidor\u002Fagente ↔ sistema comercial de comercios\u003C\u002Ftd>\u003Ctd class=\"border border-gray-300 px-4 py-2\">Capacidades de producto\u002Fcarrito\u002Fpago (checkout)\u002Fcumplimiento\u002Fpedidos\u003C\u002Ftd>\u003Ctd class=\"border border-gray-300 px-4 py-2\">Comunicación de propósito general entre agentes\u003C\u002Ftd>\u003C\u002Ftr>\u003Ctr>\u003Ctd class=\"border border-gray-300 px-4 py-2\">AP2\u003C\u002Ftd>\u003Ctd class=\"border border-gray-300 px-4 py-2\">Intención del usuario\u002Fagente ↔ autorización de pago\u003C\u002Ftd>\u003Ctd class=\"border border-gray-300 px-4 py-2\">Mandatos, restricciones de aprobación y autoridad auditable de pagos liderados por agentes\u003C\u002Ftd>\u003Ctd class=\"border border-gray-300 px-4 py-2\">Descubrimiento de productos o transporte genérico de pasarelas de pago\u003C\u002Ftd>\u003C\u002Ftr>\u003Ctr>\u003Ctd class=\"border border-gray-300 px-4 py-2\">A2UI\u003C\u002Ftd>\u003Ctd class=\"border border-gray-300 px-4 py-2\">Agente ↔ host de interfaz de usuario\u003C\u002Ftd>\u003Ctd class=\"border border-gray-300 px-4 py-2\">Intención declarativa de UI renderizada mediante componentes nativos confiables\u003C\u002Ftd>\u003Ctd class=\"border border-gray-300 px-4 py-2\">Código frontend remoto arbitrario o delegación de tareas entre agentes\u003C\u002Ftd>\u003C\u002Ftr>\u003C\u002Ftbody>\u003C\u002Ftable>\u003C\u002Fdiv>\n\u003Caside class=\"editorjs-callout editorjs-callout--success my-6 rounded-xl border p-5 border-emerald-300 bg-emerald-50 dark:border-emerald-900 dark:bg-emerald-950\u002F20\" role=\"note\">\u003Cstrong class=\"block mb-2 text-gray-900 dark:text-gray-100\">Un único flujo de trabajo puede utilizar los cinco\u003C\u002Fstrong>\u003Cdiv class=\"text-gray-700 dark:text-gray-200\">Un agente puede usar <strong>MCP<\u002Fstrong> para inspeccionar el inventario, <strong>A2A<\u002Fstrong> para consultar la disponibilidad a un agente proveedor, <strong>UCP<\u002Fstrong> para estructurar una transacción comercial, <strong>AP2<\u002Fstrong> para certificar la autorización de gasto y <strong>A2UI<\u002Fstrong> para renderizar una interfaz de aprobación nativa para el usuario.\u003C\u002Fdiv>\u003C\u002Faside>\n\u003Ch2 id=\"section-12\">1. MCP: conectar el agente a las capacidades\u003C\u002Fh2>\n\u003Cp>El Model Context Protocol es un estándar abierto para conectar aplicaciones de IA a sistemas externos donde residen herramientas, datos y recursos reutilizables. Un servidor expone capacidades; un host de MCP se conecta a ese servidor y pone dichas capacidades a disposición del modelo o aplicación.\u003C\u002Fp>\n\u003Cp>La documentación actual de MCP TypeScript v2 describe el protocolo exactamente en esos términos: los servidores exponen herramientas, recursos y prompts, mientras que los hosts, como entornos de desarrollo o aplicaciones personalizadas, se conectan a ellos. Esto convierte a MCP principalmente en un protocolo de integración de capacidades.\u003C\u002Fp>\n\u003Ch3 id=\"section-15\">Utilice MCP cuando\u003C\u002Fh3>\n\u003Cul>\u003Cli>Una aplicación de IA necesite acceso estandarizado a herramientas o API.\u003C\u002Fli>\u003Cli>Desea que un único servidor de capacidades funcione con múltiples hosts de IA compatibles.\u003C\u002Fli>\u003Cli>Necesite acceso estructurado a datos o recursos sin codificar de forma fija cada integración en cada agente.\u003C\u002Fli>\u003Cli>El sistema externo sea un proveedor de capacidades, no un agente autónomo equivalente.\u003C\u002Fli>\u003C\u002Ful>\n\u003Caside class=\"editorjs-callout editorjs-callout--warning my-6 rounded-xl border p-5 border-amber-300 bg-amber-50 dark:border-amber-900 dark:bg-amber-950\u002F20\" role=\"note\">\u003Cstrong class=\"block mb-2 text-gray-900 dark:text-gray-100\">MCP no equivale automáticamente a A2A\u003C\u002Fstrong>\u003Cdiv class=\"text-gray-700 dark:text-gray-200\">Un servidor MCP puede exponer funciones avanzadas, pero eso no lo convierte en un agente independiente con su propio ciclo de vida de tareas, semántica de descubrimiento y razonamiento interno opaco. La invocación de herramientas y la colaboración entre agentes son contratos distintos.\u003C\u002Fdiv>\u003C\u002Faside>\n\u003Ch2 id=\"section-18\">2. A2A: conectar agentes independientes\u003C\u002Fh2>\n\u003Cp>Agent2Agent (A2A) está diseñado para la comunicación entre sistemas de agentes independientes y potencialmente opacos. Su especificación actual v1.0 se centra en el descubrimiento de capacidades, mensajería, tareas, artefactos, contenido multimodal y colaboración de larga duración sin requerir que un agente exponga sus herramientas internas, memoria o implementación a otro.\u003C\u002Fp>\n\u003Cp>Esa opacidad es el límite importante. El agente invocador no necesita saber si el agente remoto utiliza internamente MCP, herramientas personalizadas, un planificador propietario, otro proveedor de modelos o escalamiento humano. Necesita un contrato para descubrir capacidades y delegar trabajo.\u003C\u002Fp>\n\u003Cp>A2A v1.0 también estandariza la negociación de versiones y admite múltiples enlaces alrededor de un modelo de datos común. Su mecanismo publicado de Agent Card ofrece a los clientes un punto de descubrimiento estándar para las capacidades, protocolos admitidos, requisitos de autenticación y habilidades de un agente.\u003C\u002Fp>\n\u003Ch3 id=\"section-22\">Use A2A cuando\u003C\u002Fh3>\n\u003Cul>\u003Cli>Un agente autónomo necesita delegar trabajo en otro agente autónomo.\u003C\u002Fli>\u003Cli>El sistema remoto debe permanecer opaco detrás de un contrato de capacidades.\u003C\u002Fli>\u003Cli>Las tareas pueden ser de larga duración, asíncronas o requerir interacción humana en el bucle (human-in-the-loop).\u003C\u002Fli>\u003Cli>Los agentes están construidos con diferentes frameworks, lenguajes, proveedores o pertenencia organizativa.\u003C\u002Fli>\u003C\u002Ful>\n\u003Ch2 id=\"section-24\">MCP vs. A2A: integración vertical frente a colaboración horizontal\u003C\u002Fh2>\n\u003Csection class=\"editorjs-comparison my-6\">\u003Ch3 class=\"mb-3 text-lg font-semibold\">MCP y A2A resuelven diferentes problemas de interoperabilidad\u003C\u002Fh3>\u003Cdiv class=\"overflow-x-auto\">\u003Ctable class=\"w-full border-collapse\">\u003Cthead>\u003Ctr>\u003Cth class=\"border border-gray-300 bg-gray-50 px-4 py-3 text-left dark:border-gray-700 dark:bg-gray-900\">\u003C\u002Fth>\u003Cth class=\"border border-gray-300 bg-gray-50 px-4 py-3 text-left font-semibold dark:border-gray-700 dark:bg-gray-900\">Dimensión\u003C\u002Fth>\u003Cth class=\"border border-gray-300 bg-gray-50 px-4 py-3 text-left font-semibold dark:border-gray-700 dark:bg-gray-900\">MCP\u003C\u002Fth>\u003Cth class=\"border border-gray-300 bg-gray-50 px-4 py-3 text-left font-semibold dark:border-gray-700 dark:bg-gray-900\">A2A\u003C\u002Fth>\u003C\u002Ftr>\u003C\u002Fthead>\u003Ctbody>\u003Ctr>\u003Cth class=\"border border-gray-300 px-4 py-3 text-left font-semibold dark:border-gray-700\">Relación\u003C\u002Fth>\u003Ctd class=\"border border-gray-300 px-4 py-3 dark:border-gray-700\">\u003C\u002Ftd>\u003Ctd class=\"border border-gray-300 px-4 py-3 dark:border-gray-700\">\u003C\u002Ftd>\u003Ctd class=\"border border-gray-300 px-4 py-3 dark:border-gray-700\">\u003C\u002Ftd>\u003C\u002Ftr>\u003Ctr>\u003Cth class=\"border border-gray-300 px-4 py-3 text-left font-semibold dark:border-gray-700\">Abstracción\u003C\u002Fth>\u003Ctd class=\"border border-gray-300 px-4 py-3 dark:border-gray-700\">\u003C\u002Ftd>\u003Ctd class=\"border border-gray-300 px-4 py-3 dark:border-gray-700\">\u003C\u002Ftd>\u003Ctd class=\"border border-gray-300 px-4 py-3 dark:border-gray-700\">\u003C\u002Ftd>\u003C\u002Ftr>\u003Ctr>\u003Cth class=\"border border-gray-300 px-4 py-3 text-left font-semibold dark:border-gray-700\">Opacidad interna\u003C\u002Fth>\u003Ctd class=\"border border-gray-300 px-4 py-3 dark:border-gray-700\">\u003C\u002Ftd>\u003Ctd class=\"border border-gray-300 px-4 py-3 dark:border-gray-700\">\u003C\u002Ftd>\u003Ctd class=\"border border-gray-300 px-4 py-3 dark:border-gray-700\">\u003C\u002Ftd>\u003C\u002Ftr>\u003Ctr>\u003Cth class=\"border border-gray-300 px-4 py-3 text-left font-semibold dark:border-gray-700\">Trabajo de larga duración\u003C\u002Fth>\u003Ctd class=\"border border-gray-300 px-4 py-3 dark:border-gray-700\">\u003C\u002Ftd>\u003Ctd class=\"border border-gray-300 px-4 py-3 dark:border-gray-700\">\u003C\u002Ftd>\u003Ctd class=\"border border-gray-300 px-4 py-3 dark:border-gray-700\">\u003C\u002Ftd>\u003C\u002Ftr>\u003C\u002Ftbody>\u003C\u002Ftable>\u003C\u002Fdiv>\u003C\u002Fsection>\n\u003Cp>El propio proyecto A2A describe ahora la distinción como horizontal frente a vertical: MCP conecta agentes con herramientas y bases de datos internas, mientras que A2A permite la colaboración peer-to-peer entre sistemas de agentes.\u003C\u002Fp>\n\u003Ch2 id=\"section-27\">3. UCP: estandarizar el comercio agéntico\u003C\u002Fh2>\n\u003Cp>El Universal Commerce Protocol no es un protocolo de agentes genérico. Estandariza los recorridos comerciales entre interfaces de consumidores, comerciantes y proveedores de pagos. La implementación de Google ya admite capacidades como la creación de carritos, el checkout, el cumplimiento de pedidos y el ciclo de vida de los pedidos a través de perfiles y API versionadas.\u003C\u002Fp>\n\u003Cp>Un comerciante puede publicar un perfil de UCP en \u002F.well-known\u002Fucp describiendo servicios, versiones de protocolo y capacidades. Ese patrón de descubrimiento es importante porque una interfaz agéntica no debería necesitar un contrato de checkout a medida para cada comerciante.\u003C\u002Fp>\n\u003Cp>UCP también es intencionalmente componible. La descripción técnica general de Google indica que puede integrarse a través de API, A2A y MCP, y es compatible con AP2 para la autorización de pagos agénticos.\u003C\u002Fp>\n\u003Ch3 id=\"section-31\">Use UCP cuando\u003C\u002Fh3>\n\u003Cul>\u003Cli>El flujo de trabajo involucra productos de comerciantes, carritos, checkout, cumplimiento o ciclo de vida del pedido.\u003C\u002Fli>\u003Cli>Está construyendo una interfaz para comerciantes que debe funcionar con experiencias de compra agénticas.\u003C\u002Fli>\u003Cli>La integración necesita semántica específica de comercio en lugar de llamadas a herramientas genéricas.\u003C\u002Fli>\u003Cli>Desea un contrato de comercio interoperable que pueda coexistir con MCP, A2A y protocolos de pago.\u003C\u002Fli>\u003C\u002Ful>\n\u003Ch2 id=\"section-33\">4. AP2: demostrar que el agente tenía autorización para gastar\u003C\u002Fh2>\n\u003Cp>El comercio agéntico introduce un problema que los flujos de checkout ordinarios no tenían que resolver de la misma manera: un agente puede realizar transacciones cuando el humano no está haciendo clic en el botón final en tiempo real. El Agent Payments Protocol (AP2) aborda la autorización, la autenticidad y la rendición de cuentas para los pagos dirigidos por agentes.\u003C\u002Fp>\n\u003Cp>La guía de protocolos de Google de 2026 describe AP2 a través de mandatos tipados que capturan la intención del usuario, las restricciones de gasto y la transacción específica que se autoriza. AP2 puede funcionar como una extensión junto a UCP: UCP describe la transacción comercial, mientras que AP2 proporciona pruebas de que el agente tenía autorización para realizar el pago.\u003C\u002Fp>\n\u003Cp>Esta distinción es importante. Un protocolo de checkout puede indicarle a un comerciante qué debe comprarse. Por sí solo no demuestra quién autorizó al agente a gastar, bajo qué límite, para qué comerciante, durante cuánto tiempo o si el carrito final se mantuvo dentro de esa autorización.\u003C\u002Fp>\n\u003Ch2 id=\"section-37\">UCP vs. AP2: semántica de transacciones frente a autoridad\u003C\u002Fh2>\n\u003Cdiv class=\"overflow-x-auto\">\u003Ctable class=\"w-full border-collapse\">\u003Cthead>\u003Ctr>\u003Cth class=\"border border-gray-300 px-4 py-2 text-left font-semibold\">Pregunta\u003C\u002Fth>\u003Cth class=\"border border-gray-300 px-4 py-2 text-left font-semibold\">UCP\u003C\u002Fth>\u003Cth class=\"border border-gray-300 px-4 py-2 text-left font-semibold\">AP2\u003C\u002Fth>\u003C\u002Ftr>\u003C\u002Fthead>\u003Ctbody>\u003Ctr>\u003Ctd class=\"border border-gray-300 px-4 py-2\">¿Qué se está comprando?\u003C\u002Ftd>\u003Ctd class=\"border border-gray-300 px-4 py-2\">Artículos de comercio, carrito, semántica de checkout y cumplimiento\u003C\u002Ftd>\u003Ctd class=\"border border-gray-300 px-4 py-2\">Hace referencia al contexto de la transacción autorizada\u003C\u002Ftd>\u003C\u002Ftr>\u003Ctr>\u003Ctd class=\"border border-gray-300 px-4 py-2\">¿Quién puede autorizarlo?\u003C\u002Ftd>\u003Ctd class=\"border border-gray-300 px-4 py-2\">No es la responsabilidad principal del protocolo\u003C\u002Ftd>\u003Ctd class=\"border border-gray-300 px-4 py-2\">Modelo explícito de autoridad y mandatos de agente\u002Fusuario\u003C\u002Ftd>\u003C\u002Ftr>\u003Ctr>\u003Ctd class=\"border border-gray-300 px-4 py-2\">¿Qué restricciones de gasto aplican?\u003C\u002Ftd>\u003Ctd class=\"border border-gray-300 px-4 py-2\">El flujo de comercio puede contener totales y datos de checkout\u003C\u002Ftd>\u003Ctd class=\"border border-gray-300 px-4 py-2\">Límites de autorización y restricciones de intención\u003C\u002Ftd>\u003C\u002Ftr>\u003Ctr>\u003Ctd class=\"border border-gray-300 px-4 py-2\">¿Cómo se audita la transacción?\u003C\u002Ftd>\u003Ctd class=\"border border-gray-300 px-4 py-2\">Ciclo de vida del pedido y del comercio\u003C\u002Ftd>\u003Ctd class=\"border border-gray-300 px-4 py-2\">Rastro de autorización criptográfico \u002F verificable mediante mandatos y recibos\u003C\u002Ftd>\u003C\u002Ftr>\u003Ctr>\u003Ctd class=\"border border-gray-300 px-4 py-2\">¿Pueden trabajar juntos?\u003C\u002Ftd>\u003Ctd class=\"border border-gray-300 px-4 py-2\">Sí\u003C\u002Ftd>\u003Ctd class=\"border border-gray-300 px-4 py-2\">Sí: AP2 puede ampliar los flujos de comercio agéntico\u003C\u002Ftd>\u003C\u002Ftr>\u003C\u002Ftbody>\u003C\u002Ftable>\u003C\u002Fdiv>\n\u003Ch2 id=\"section-39\">5. A2UI: permita que los agentes describan interfaces sin adueñarse de su frontend\u003C\u002Fh2>\n\u003Cp>Agent-to-User Interface (A2UI) aborda otro límite: cómo un agente remoto o local comunica una interfaz interactiva enriquecida a una aplicación anfitriona. En lugar de enviar HTML, CSS y JavaScript arbitrarios, A2UI utiliza datos declarativos que el anfitrión renderiza a través de su propio catálogo de componentes de confianza.\u003C\u002Fp>\n\u003Cp>Esto preserva el sistema de diseño y el modelo de seguridad de la aplicación anfitriona, al tiempo que permite que un agente solicite interfaces dinámicas. A2UI v0.9 enfatiza específicamente la intención de interfaz de usuario independiente del framework y las actualizaciones en streaming a través de la web, móviles y otros clientes.\u003C\u002Fp>\n\u003Cp>El trabajo posterior de Google sobre A2UI + MCP Apps también demuestra que estos modelos de interfaz de usuario no son necesariamente mutuamente excluyentes. La interfaz de usuario nativa declarativa y las experiencias más ricas de aplicaciones integradas pueden coexistir según la tarea.\u003C\u002Fp>\n\u003Ch3 id=\"section-43\">Utilice A2UI cuando\u003C\u002Fh3>\n\u003Cul>\u003Cli>Un agente remoto necesita solicitar formularios, tarjetas, controles u otra interfaz interactiva.\u003C\u002Fli>\u003Cli>El anfitrión debe preservar sus componentes nativos, estilo y límite de seguridad.\u003C\u002Fli>\u003Cli>No desea que los agentes remotos envíen código frontend ejecutable arbitrario.\u003C\u002Fli>\u003Cli>La misma intención de interfaz definida por el agente debe funcionar en diferentes frameworks de cliente.\u003C\u002Fli>\u003C\u002Ful>\n\u003Ch2 id=\"section-45\">La prueba de selección de protocolos\u003C\u002Fh2>\n\u003Cp>No empiece por el acrónimo. Empiece por la relación que necesita interoperabilidad.\u003C\u002Fp>\n\u003Csection class=\"editorjs-process my-6\">\u003Ch3 class=\"mb-3 text-lg font-semibold\">Elija el protocolo según el límite\u003C\u002Fh3>\u003Cdiv class=\"grid grid-cols-1 md:grid-cols-2 xl:grid-cols-3 gap-4\">\u003Cdiv class=\"editorjs-process__step min-w-0  rounded-xl border border-gray-200 dark:border-gray-700 p-4\">\u003Cdiv class=\"text-xs font-semibold text-gray-500 dark:text-gray-400\">1\u003C\u002Fdiv>\u003Cdiv class=\"mt-1 font-semibold text-gray-900 dark:text-gray-100\">1. Identifique las dos partes independientes\u003C\u002Fdiv>\u003Cdiv class=\"mt-1 text-sm text-gray-600 dark:text-gray-300\">¿Se trata de IA a herramienta, agente a agente, agente a comerciante, agente a autoridad de pago o agente a interfaz de usuario?\u003C\u002Fdiv>\u003C\u002Fdiv>\u003Cdiv class=\"editorjs-process__step min-w-0  rounded-xl border border-gray-200 dark:border-gray-700 p-4\">\u003Cdiv class=\"text-xs font-semibold text-gray-500 dark:text-gray-400\">2\u003C\u002Fdiv>\u003Cdiv class=\"mt-1 font-semibold text-gray-900 dark:text-gray-100\">2. Identifique el objeto compartido\u003C\u002Fdiv>\u003Cdiv class=\"mt-1 text-sm text-gray-600 dark:text-gray-300\">¿El contrato trata sobre una llamada a herramienta, tarea, carrito, mandato de pago, artefacto o descripción de interfaz de usuario?\u003C\u002Fdiv>\u003C\u002Fdiv>\u003Cdiv class=\"editorjs-process__step min-w-0  rounded-xl border border-gray-200 dark:border-gray-700 p-4\">\u003Cdiv class=\"text-xs font-semibold text-gray-500 dark:text-gray-400\">3\u003C\u002Fdiv>\u003Cdiv class=\"mt-1 font-semibold text-gray-900 dark:text-gray-100\">3. Compruebe si ya existe un protocolo de dominio\u003C\u002Fdiv>\u003Cdiv class=\"mt-1 text-sm text-gray-600 dark:text-gray-300\">Prefiera la semántica de comercio o pago cuando el problema sea de comercio o autorización, en lugar de codificar todo como herramientas genéricas.\u003C\u002Fdiv>\u003C\u002Fdiv>\u003Cdiv class=\"editorjs-process__step min-w-0  rounded-xl border border-gray-200 dark:border-gray-700 p-4\">\u003Cdiv class=\"text-xs font-semibold text-gray-500 dark:text-gray-400\">4\u003C\u002Fdiv>\u003Cdiv class=\"mt-1 font-semibold text-gray-900 dark:text-gray-100\">4. Mantenga los aspectos internos locales como locales\u003C\u002Fdiv>\u003Cdiv class=\"mt-1 text-sm text-gray-600 dark:text-gray-300\">No exponga un agente completo como herramientas MCP si la parte remota solo necesita una capacidad A2A, y no haga que un agente remoto sea responsable del tiempo de ejecución de su interfaz de usuario.\u003C\u002Fdiv>\u003C\u002Fdiv>\u003Cdiv class=\"editorjs-process__step min-w-0  rounded-xl border border-gray-200 dark:border-gray-700 p-4\">\u003Cdiv class=\"text-xs font-semibold text-gray-500 dark:text-gray-400\">5\u003C\u002Fdiv>\u003Cdiv class=\"mt-1 font-semibold text-gray-900 dark:text-gray-100\">5. Componga protocolos cuando el flujo de trabajo cruce límites\u003C\u002Fdiv>\u003Cdiv class=\"mt-1 text-sm text-gray-600 dark:text-gray-300\">Un flujo de trabajo puede cruzar legítimamente contratos de herramientas, agentes, comercio, pagos e interfaz de usuario.\u003C\u002Fdiv>\u003C\u002Fdiv>\u003Cdiv class=\"editorjs-process__step min-w-0  rounded-xl border border-gray-200 dark:border-gray-700 p-4\">\u003Cdiv class=\"text-xs font-semibold text-gray-500 dark:text-gray-400\">6\u003C\u002Fdiv>\u003Cdiv class=\"mt-1 font-semibold text-gray-900 dark:text-gray-100\">6. Controle las versiones de cada contrato de forma independiente\u003C\u002Fdiv>\u003Cdiv class=\"mt-1 text-sm text-gray-600 dark:text-gray-300\">Las versiones de los protocolos evolucionan a diferentes velocidades; no vincule cada integración a una única versión monolítica de la aplicación.\u003C\u002Fdiv>\u003C\u002Fdiv>\u003Cdiv class=\"editorjs-process__step min-w-0  rounded-xl border border-gray-200 dark:border-gray-700 p-4\">\u003Cdiv class=\"text-xs font-semibold text-gray-500 dark:text-gray-400\">7\u003C\u002Fdiv>\u003Cdiv class=\"mt-1 font-semibold text-gray-900 dark:text-gray-100\">7. Preserve la autorización en cada límite\u003C\u002Fdiv>\u003Cdiv class=\"mt-1 text-sm text-gray-600 dark:text-gray-300\">La interoperabilidad no reemplaza los permisos de producto, la autorización de herramientas, la autoridad de pago ni la política de acceso a datos.\u003C\u002Fdiv>\u003C\u002Fdiv>\u003C\u002Fdiv>\u003C\u002Fsection>\n\u003Ch2 id=\"section-48\">Un flujo de trabajo multiprotocolo realista\u003C\u002Fh2>\n\u003Csection class=\"editorjs-process my-6\">\u003Ch3 class=\"mb-3 text-lg font-semibold\">Ejemplo: un flujo de trabajo de aprovisionamiento autónomo\u003C\u002Fh3>\u003Cdiv class=\"grid grid-cols-1 md:grid-cols-2 xl:grid-cols-3 gap-4\">\u003Cdiv class=\"editorjs-process__step min-w-0  rounded-xl border border-gray-200 dark:border-gray-700 p-4\">\u003Cdiv class=\"text-xs font-semibold text-gray-500 dark:text-gray-400\">1\u003C\u002Fdiv>\u003Cdiv class=\"mt-1 font-semibold text-gray-900 dark:text-gray-100\">1. Inspeccionar el stock interno con MCP\u003C\u002Fdiv>\u003Cdiv class=\"mt-1 text-sm text-gray-600 dark:text-gray-300\">El agente de compras llama a las capacidades de inventario y previsión expuestas por los servidores MCP internos.\u003C\u002Fdiv>\u003C\u002Fdiv>\u003Cdiv class=\"editorjs-process__step min-w-0  rounded-xl border border-gray-200 dark:border-gray-700 p-4\">\u003Cdiv class=\"text-xs font-semibold text-gray-500 dark:text-gray-400\">2\u003C\u002Fdiv>\u003Cdiv class=\"mt-1 font-semibold text-gray-900 dark:text-gray-100\">2. Descubrir un agente proveedor con A2A\u003C\u002Fdiv>\u003Cdiv class=\"mt-1 text-sm text-gray-600 dark:text-gray-300\">El agente lee la Agent Card del proveedor y delega una tarea de disponibilidad y tiempo de entrega.\u003C\u002Fdiv>\u003C\u002Fdiv>\u003Cdiv class=\"editorjs-process__step min-w-0  rounded-xl border border-gray-200 dark:border-gray-700 p-4\">\u003Cdiv class=\"text-xs font-semibold text-gray-500 dark:text-gray-400\">3\u003C\u002Fdiv>\u003Cdiv class=\"mt-1 font-semibold text-gray-900 dark:text-gray-100\">3. Negociar el objeto comercial con UCP\u003C\u002Fdiv>\u003Cdiv class=\"mt-1 text-sm text-gray-600 dark:text-gray-300\">La superficie del proveedor o comerciante devuelve información estructurada sobre el carrito, el checkout y el cumplimiento.\u003C\u002Fdiv>\u003C\u002Fdiv>\u003Cdiv class=\"editorjs-process__step min-w-0  rounded-xl border border-gray-200 dark:border-gray-700 p-4\">\u003Cdiv class=\"text-xs font-semibold text-gray-500 dark:text-gray-400\">4\u003C\u002Fdiv>\u003Cdiv class=\"mt-1 font-semibold text-gray-900 dark:text-gray-100\">4. Verificar la autoridad de gasto con AP2\u003C\u002Fdiv>\u003Cdiv class=\"mt-1 text-sm text-gray-600 dark:text-gray-300\">La compra se compara con el mandato firmado del usuario o de la organización, las restricciones del comerciante y los límites de gasto.\u003C\u002Fdiv>\u003C\u002Fdiv>\u003Cdiv class=\"editorjs-process__step min-w-0  rounded-xl border border-gray-200 dark:border-gray-700 p-4\">\u003Cdiv class=\"text-xs font-semibold text-gray-500 dark:text-gray-400\">5\u003C\u002Fdiv>\u003Cdiv class=\"mt-1 font-semibold text-gray-900 dark:text-gray-100\">5. Solicitar aprobación a través de A2UI\u003C\u002Fdiv>\u003Cdiv class=\"mt-1 text-sm text-gray-600 dark:text-gray-300\">Si se requiere aprobación humana, el agente envía una intención declarativa de interfaz de usuario y el anfitrión renderiza la experiencia de aprobación utilizando componentes nativos confiables.\u003C\u002Fdiv>\u003C\u002Fdiv>\u003Cdiv class=\"editorjs-process__step min-w-0  rounded-xl border border-gray-200 dark:border-gray-700 p-4\">\u003Cdiv class=\"text-xs font-semibold text-gray-500 dark:text-gray-400\">6\u003C\u002Fdiv>\u003Cdiv class=\"mt-1 font-semibold text-gray-900 dark:text-gray-100\">6. Completar y auditar\u003C\u002Fdiv>\u003Cdiv class=\"mt-1 text-sm text-gray-600 dark:text-gray-300\">El estado del comercio, la autorización de pago, la evidencia de la tarea del agente y los registros de auditoría de la aplicación siguen siendo rastreables a través de sus respectivos límites.\u003C\u002Fdiv>\u003C\u002Fdiv>\u003C\u002Fdiv>\u003C\u002Fsection>\n\u003Ch2 id=\"section-50\">Por qué es poco probable que un protocolo universal de agentes los reemplace a todos\u003C\u002Fh2>\n\u003Cp>Un protocolo universal suena más sencillo hasta que debe codificar la semántica de cada dominio. El descubrimiento de herramientas, la colaboración prolongada entre agentes, el checkout, la autorización de pagos y la interfaz de usuario nativa tienen diferentes requisitos de ciclo de vida, seguridad y corrección.\u003C\u002Fp>\n\u003Cp>La propia web evolucionó mediante protocolos por capas en lugar de un único formato de mensaje para cada problema. El stack agéntico emergente parece avanzar en la misma dirección: primitivas horizontales comunes, contratos de dominio especializados y descubrimiento\u002Fversionado explícitos.\u003C\u002Fp>\n\u003Cp>Por lo tanto, el desafío arquitectónico pasa de «¿qué protocolo gana?» a con qué claridad se componen los protocolos sin duplicar la identidad, la autorización, el estado y la semántica de auditoría.\u003C\u002Fp>\n\u003Ch2 id=\"section-54\">La composición de protocolos genera nuevos modos de fallo\u003C\u002Fh2>\n\u003Cdiv class=\"overflow-x-auto\">\u003Ctable class=\"w-full border-collapse\">\u003Cthead>\u003Ctr>\u003Cth class=\"border border-gray-300 px-4 py-2 text-left font-semibold\">Modo de fallo\u003C\u002Fth>\u003Cth class=\"border border-gray-300 px-4 py-2 text-left font-semibold\">Qué sucede\u003C\u002Fth>\u003Cth class=\"border border-gray-300 px-4 py-2 text-left font-semibold\">Control de arquitectura\u003C\u002Fth>\u003C\u002Ftr>\u003C\u002Fthead>\u003Ctbody>\u003Ctr>\u003Ctd class=\"border border-gray-300 px-4 py-2\">Fuga de autoridad\u003C\u002Ftd>\u003Ctd class=\"border border-gray-300 px-4 py-2\">Una capacidad válida de herramienta o agente se trata como permiso para realizar una acción de negocio\u003C\u002Ftd>\u003Ctd class=\"border border-gray-300 px-4 py-2\">Mantener la autorización del producto independiente del descubrimiento de capacidades del protocolo\u003C\u002Ftd>\u003C\u002Ftr>\u003Ctr>\u003Ctd class=\"border border-gray-300 px-4 py-2\">Discrepancia de identidad\u003C\u002Ftd>\u003Ctd class=\"border border-gray-300 px-4 py-2\">La identidad del host de MCP, la identidad del agente A2A y la identidad de comercio\u002Fpago hacen referencia a diferentes entidades principales\u003C\u002Ftd>\u003Ctd class=\"border border-gray-300 px-4 py-2\">Definir un mapeo explícito de entidades principales a través de los límites\u003C\u002Ftd>\u003C\u002Ftr>\u003Ctr>\u003Ctd class=\"border border-gray-300 px-4 py-2\">Desfase de versiones\u003C\u002Ftd>\u003Ctd class=\"border border-gray-300 px-4 py-2\">Un protocolo se actualiza mientras que los adaptadores dependientes asumen una semántica anterior\u003C\u002Ftd>\u003Ctd class=\"border border-gray-300 px-4 py-2\">Negociar y fijar versiones de protocolo de forma independiente\u003C\u002Ftd>\u003C\u002Ftr>\u003Ctr>\u003Ctd class=\"border border-gray-300 px-4 py-2\">Duplicación de estado\u003C\u002Ftd>\u003Ctd class=\"border border-gray-300 px-4 py-2\">El mismo estado de carrito, tarea o aprobación se copia en varias capas de protocolo\u003C\u002Ftd>\u003Ctd class=\"border border-gray-300 px-4 py-2\">Definir un único propietario con autoridad por objeto de dominio\u003C\u002Ftd>\u003C\u002Ftr>\u003Ctr>\u003Ctd class=\"border border-gray-300 px-4 py-2\">Fragmentación de auditoría\u003C\u002Ftd>\u003Ctd class=\"border border-gray-300 px-4 py-2\">Las trazas de herramientas, las tareas de agentes y la evidencia de checkout y pagos no se pueden vincular\u003C\u002Ftd>\u003Ctd class=\"border border-gray-300 px-4 py-2\">Transmitir identificadores de correlación e identificadores de dominio estables a través de los límites del protocolo\u003C\u002Ftd>\u003C\u002Ftr>\u003Ctr>\u003Ctd class=\"border border-gray-300 px-4 py-2\">Tunelización semántica\u003C\u002Ftd>\u003Ctd class=\"border border-gray-300 px-4 py-2\">Todo se fuerza a través de un protocolo genérico como JSON opaco\u003C\u002Ftd>\u003Ctd class=\"border border-gray-300 px-4 py-2\">Utilizar protocolos de dominio donde su semántica mejore materialmente la corrección\u003C\u002Ftd>\u003C\u002Ftr>\u003C\u002Ftbody>\u003C\u002Ftable>\u003C\u002Fdiv>\n\u003Ch2 id=\"section-56\">La elección del protocolo no reemplaza la arquitectura de la aplicación\u003C\u002Fh2>\n\u003Cp>Los estándares abiertos reducen el acoplamiento de integración, pero no deciden su modelo de dominio, política de autorización, fuente de verdad, estrategia de reintentos o criterios de aceptación. Una herramienta MCP aún puede exponer la capacidad incorrecta. Un agente A2A aún puede devolver un artefacto erróneo. Un checkout de UCP aún puede contener datos desactualizados del comerciante. La lógica de la aplicación aún puede aplicar incorrectamente un mandato AP2.\u003C\u002Fp>\n\u003Cp>Trate los protocolos como contratos entre componentes que evolucionan de forma independiente. Mantenga la verdad del dominio y la política consecuente en la capa de aplicación a la que pertenecen, y luego use protocolos para hacer que los límites sean interoperables.\u003C\u002Fp>\n\u003Caside class=\"editorjs-referral my-6\">\u003Ca href=\"https:\u002F\u002Fstajic.de\u002Fes\u002Fblog\u002Fmanaged-agent-harness-vs-self-hosted-agent-loop-what-you-gain-what-you-lose\" class=\"flex flex-col sm:flex-row gap-4 rounded-xl border border-gray-200 dark:border-gray-700 p-4 transition hover:border-primary-500\">\u003Cdiv class=\"min-w-0 flex-1\">\u003Cstrong class=\"block text-lg text-gray-900 dark:text-gray-100\">Harness de agentes gestionado vs. bucle de agentes autohospedado: lo que gana, lo que pierde\u003C\u002Fstrong>\u003Cp class=\"mt-2 text-sm text-gray-600 dark:text-gray-300\">Los límites de protocolo resuelven la interoperabilidad. Los límites de tiempo de ejecución resuelven quién opera el harness, el entorno de ejecución y el plano de control de la aplicación.\u003C\u002Fp>\u003Cspan class=\"mt-3 inline-flex text-sm font-medium text-primary-600 dark:text-primary-400\">Leer la guía de arquitectura de tiempo de ejecución →\u003C\u002Fspan>\u003C\u002Fdiv>\u003C\u002Fa>\u003C\u002Faside>\n\u003Ch2 id=\"section-60\">¿Qué cambiaría esta respuesta?\u003C\u002Fh2>\n\u003Cp>La pila tecnológica cambia si los protocolos convergen, si un estándar absorbe formalmente a otro o si los proveedores estandarizan una capa compartida de identidad y autorización a través de múltiples límites. UCP ya demuestra la composición al admitir APIs, A2A y MCP, y al integrarse con AP2 en lugar de reemplazarlos.\u003C\u002Fp>\n\u003Cp>La respuesta también varía según el alcance de la aplicación. Un agente interno pequeño puede necesitar únicamente MCP. Un flujo de trabajo entre múltiples empresas puede necesitar A2A. Un comerciante puede necesitar UCP sin A2UI. Un agente de compras delegado puede necesitarlos todos. Utilice el conjunto de protocolos más pequeño que represente los límites reales sin aplanar la semántica de dominio.\u003C\u002Fp>\n\u003Ch2 id=\"section-63\">Limitaciones\u003C\u002Fh2>\n\u003Cp>Los protocolos analizados aquí se encuentran en diferentes niveles de madurez y tienen distintos modelos de gobernanza. A2A ha alcanzado una especificación v1.0 estable, mientras que otros estándares continúan evolucionando rápidamente. La adopción en el ecosistema también es desigual entre proveedores y frameworks.\u003C\u002Fp>\n\u003Cp>Este artículo se enfoca en la responsabilidad arquitectónica más que en la exhaustividad de la implementación. Los métodos de autenticación específicos, los enlaces de transporte, los esquemas y los mecanismos de extensión deben tomarse de la especificación vigente de cada protocolo.\u003C\u002Fp>\n\u003Ch2 id=\"section-66\">Conclusión\u003C\u002Fh2>\n\u003Cp>MCP, A2A, UCP, AP2 y A2UI cobran más sentido cuando se consideran protocolos para diferentes relaciones, y no cinco intentos competitivos de estandarizar los “agentes”.\u003C\u002Fp>\n\u003Cp>MCP expone capacidades. A2A coordina agentes independientes. UCP otorga al comercio su propio contrato legible por máquinas. AP2 añade autoridad de pago verificable. A2UI proporciona a los agentes una vía declarativa y segura hacia las interfaces de usuario. Por lo tanto, la web agéntica emergente no está reemplazando los protocolos con IA; está creando una nueva pila de protocolos en torno a la IA.\u003C\u002Fp>\n\u003Ch2 id=\"section-69\">Preguntas frecuentes\u003C\u002Fh2>\n\u003Csection class=\"editorjs-faq my-6 rounded-xl border border-gray-200 p-5 dark:border-gray-700\">\u003Ch3 class=\"mb-3 text-lg font-semibold\">MCP, A2A, UCP, AP2 y A2UI\u003C\u002Fh3>\u003Cdiv id=\"faq1\" class=\"border-t border-gray-200 py-4 first:border-t-0 dark:border-gray-700\">\u003Ch4 class=\"font-semibold text-gray-900 dark:text-gray-100\">¿Es A2A un reemplazo de MCP?\u003C\u002Fh4>\u003Cdiv class=\"mt-2 text-gray-600 dark:text-gray-300\">No. MCP estandariza principalmente la forma en que las aplicaciones de IA acceden a herramientas, recursos y datos. A2A estandariza la colaboración entre sistemas de agentes independientes. Un agente remoto puede usar MCP internamente mientras expone una interfaz A2A.\u003C\u002Fdiv>\u003C\u002Fdiv>\u003Cdiv id=\"faq2\" class=\"border-t border-gray-200 py-4 first:border-t-0 dark:border-gray-700\">\u003Ch4 class=\"font-semibold text-gray-900 dark:text-gray-100\">¿Es UCP un reemplazo de MCP en agentes de compras?\u003C\u002Fh4>\u003Cdiv class=\"mt-2 text-gray-600 dark:text-gray-300\">Generalmente no. UCP proporciona semántica específica de comercio, como carrito, checkout y cumplimiento. MCP aún puede exponer herramientas o datos del comerciante, y UCP está diseñado para coexistir con MCP y A2A.\u003C\u002Fdiv>\u003C\u002Fdiv>\u003Cdiv id=\"faq3\" class=\"border-t border-gray-200 py-4 first:border-t-0 dark:border-gray-700\">\u003Ch4 class=\"font-semibold text-gray-900 dark:text-gray-100\">¿Cuál es la diferencia entre UCP y AP2?\u003C\u002Fh4>\u003Cdiv class=\"mt-2 text-gray-600 dark:text-gray-300\">UCP estandariza las interacciones comerciales y el ciclo de vida de las transacciones. AP2 se enfoca en demostrar que un agente tenía autorización para realizar un pago bajo restricciones definidas del usuario o de la organización.\u003C\u002Fdiv>\u003C\u002Fdiv>\u003Cdiv id=\"faq4\" class=\"border-t border-gray-200 py-4 first:border-t-0 dark:border-gray-700\">\u003Ch4 class=\"font-semibold text-gray-900 dark:text-gray-100\">¿Qué problema resuelve A2UI?\u003C\u002Fh4>\u003Cdiv class=\"mt-2 text-gray-600 dark:text-gray-300\">A2UI permite que los agentes envíen intenciones de interfaz de usuario declarativas a una aplicación host, la cual renderiza la experiencia mediante componentes nativos confiables en lugar de ejecutar código frontend remoto arbitrario.\u003C\u002Fdiv>\u003C\u002Fdiv>\u003Cdiv id=\"faq5\" class=\"border-t border-gray-200 py-4 first:border-t-0 dark:border-gray-700\">\u003Ch4 class=\"font-semibold text-gray-900 dark:text-gray-100\">¿Puede una aplicación de agentes utilizar todos estos protocolos?\u003C\u002Fh4>\u003Cdiv class=\"mt-2 text-gray-600 dark:text-gray-300\">Sí. Un flujo de trabajo puede utilizar MCP para herramientas internas, A2A para delegación a agentes remotos, UCP para comercio, AP2 para autorización de pagos y A2UI para interacción humana.\u003C\u002Fdiv>\u003C\u002Fdiv>\u003Cdiv id=\"faq6\" class=\"border-t border-gray-200 py-4 first:border-t-0 dark:border-gray-700\">\u003Ch4 class=\"font-semibold text-gray-900 dark:text-gray-100\">¿Qué protocolo debería implementar primero?\u003C\u002Fh4>\u003Cdiv class=\"mt-2 text-gray-600 dark:text-gray-300\">Comience a partir del límite de interoperabilidad. Si el problema es el acceso a herramientas, evalúe MCP. Si es la colaboración entre agentes independientes, evalúe A2A. Si es de comercio, UCP. Si es la autoridad de pago delegada, AP2. Si es una interfaz de usuario portátil impulsada por agentes, A2UI.\u003C\u002Fdiv>\u003C\u002Fdiv>\u003C\u002Fsection>\n\u003Ch2 id=\"section-71\">Glosario\u003C\u002Fh2>\n\u003Csection class=\"editorjs-glossary my-6 rounded-xl border border-gray-200 dark:border-gray-700 p-5\">\u003Ch3 class=\"mb-3 text-lg font-semibold\">Términos clave sobre protocolos de agentes\u003C\u002Fh3>\u003Cdl>\u003Cdiv id=\"mcp\" class=\"border-t border-gray-200 dark:border-gray-700 py-3 first:border-t-0\">\u003Cdt class=\"font-semibold text-gray-900 dark:text-gray-100\">MCP\u003C\u002Fdt>\u003Cdd class=\"mt-1 text-gray-600 dark:text-gray-300\">Model Context Protocol, un estándar abierto para exponer herramientas, recursos y prompts desde sistemas externos hacia hosts de IA compatibles.\u003C\u002Fdd>\u003C\u002Fdiv>\u003Cdiv id=\"a2a\" class=\"border-t border-gray-200 dark:border-gray-700 py-3 first:border-t-0\">\u003Cdt class=\"font-semibold text-gray-900 dark:text-gray-100\">A2A\u003C\u002Fdt>\u003Cdd class=\"mt-1 text-gray-600 dark:text-gray-300\">Agent2Agent Protocol, un estándar abierto para descubrir y colaborar con sistemas de agentes independientes a través de mensajes, tareas y artefactos.\u003C\u002Fdd>\u003C\u002Fdiv>\u003Cdiv id=\"ucp\" class=\"border-t border-gray-200 dark:border-gray-700 py-3 first:border-t-0\">\u003Cdt class=\"font-semibold text-gray-900 dark:text-gray-100\">UCP\u003C\u002Fdt>\u003Cdd class=\"mt-1 text-gray-600 dark:text-gray-300\">Universal Commerce Protocol, un estándar abierto para recorridos de comercio agéntico interoperables entre interfaces de consumo, empresas y proveedores de pagos.\u003C\u002Fdd>\u003C\u002Fdiv>\u003Cdiv id=\"ap2\" class=\"border-t border-gray-200 dark:border-gray-700 py-3 first:border-t-0\">\u003Cdt class=\"font-semibold text-gray-900 dark:text-gray-100\">AP2\u003C\u002Fdt>\u003Cdd class=\"mt-1 text-gray-600 dark:text-gray-300\">Agent Payments Protocol, un estándar abierto para representar y verificar la autoridad, la intención y la rendición de cuentas en pagos gestionados por agentes.\u003C\u002Fdd>\u003C\u002Fdiv>\u003Cdiv id=\"a2ui\" class=\"border-t border-gray-200 dark:border-gray-700 py-3 first:border-t-0\">\u003Cdt class=\"font-semibold text-gray-900 dark:text-gray-100\">A2UI\u003C\u002Fdt>\u003Cdd class=\"mt-1 text-gray-600 dark:text-gray-300\">Agent-to-User Interface, un protocolo declarativo para permitir que los agentes soliciten una interfaz de usuario que se renderiza utilizando los componentes confiables de la aplicación host.\u003C\u002Fdd>\u003C\u002Fdiv>\u003Cdiv id=\"protocol-composition\" class=\"border-t border-gray-200 dark:border-gray-700 py-3 first:border-t-0\">\u003Cdt class=\"font-semibold text-gray-900 dark:text-gray-100\">Composición de protocolos\u003C\u002Fdt>\u003Cdd class=\"mt-1 text-gray-600 dark:text-gray-300\">El uso de múltiples protocolos en un solo flujo de trabajo, donde cada uno es responsable de un límite de interoperabilidad específico en lugar de forzar toda la semántica a través de un único contrato.\u003C\u002Fdd>\u003C\u002Fdiv>\u003C\u002Fdl>\u003C\u002Fsection>\n\u003Ch2 id=\"section-73\">Fuentes principales y lecturas adicionales\u003C\u002Fh2>\n\u003Ca href=\"https:\u002F\u002Fdevelopers.googleblog.com\u002Fdevelopers-guide-to-ai-agent-protocols\u002F\" target=\"_blank\" rel=\"noopener noreferrer\" class=\"editorjs-link-tool block border border-gray-200 dark:border-gray-700 rounded-lg p-4 transition text-gray-900 dark:text-gray-100 hover:border-primary-500 hover:bg-primary-50 dark:hover:bg-gray-900 hover:text-gray-900 dark:hover:text-gray-100\">\u003Cstrong class=\"block font-semibold\">Google Developers — Guía para desarrolladores sobre protocolos de agentes de IA\u003C\u002Fstrong>\u003Cp class=\"text-sm text-gray-600 dark:text-gray-400\">Una descripción práctica que muestra cómo MCP, A2A, UCP, AP2, A2UI y protocolos relacionados funcionan juntos en un flujo de trabajo de agente de varios pasos.\u003C\u002Fp>\u003C\u002Fa>\n\u003Ca href=\"https:\u002F\u002Fts.sdk.modelcontextprotocol.io\u002Fv2\u002F\" target=\"_blank\" rel=\"noopener noreferrer\" class=\"editorjs-link-tool block border border-gray-200 dark:border-gray-700 rounded-lg p-4 transition text-gray-900 dark:text-gray-100 hover:border-primary-500 hover:bg-primary-50 dark:hover:bg-gray-900 hover:text-gray-900 dark:hover:text-gray-100\">\u003Cstrong class=\"block font-semibold\">Model Context Protocol — TypeScript SDK v2\u003C\u002Fstrong>\u003Cp class=\"text-sm text-gray-600 dark:text-gray-400\">Documentación actual y estable del SDK que implementa la especificación MCP del 28-07-2026 y define herramientas, recursos, prompts e integración de host\u002Fservidor.\u003C\u002Fp>\u003C\u002Fa>\n\u003Ca href=\"https:\u002F\u002Fa2a-protocol.org\u002Fdev\u002Fspecification\u002F\" target=\"_blank\" rel=\"noopener noreferrer\" class=\"editorjs-link-tool block border border-gray-200 dark:border-gray-700 rounded-lg p-4 transition text-gray-900 dark:text-gray-100 hover:border-primary-500 hover:bg-primary-50 dark:hover:bg-gray-900 hover:text-gray-900 dark:hover:text-gray-100\">\u003Cstrong class=\"block font-semibold\">Protocolo A2A — Especificación v1.0\u003C\u002Fstrong>\u003Cp class=\"text-sm text-gray-600 dark:text-gray-400\">Especificación actual del protocolo A2A que abarca Agent Cards, mensajes, tareas, artefactos, enlaces y negociación de versiones.\u003C\u002Fp>\u003C\u002Fa>\n\u003Ca href=\"https:\u002F\u002Fa2a-protocol.org\u002Flatest\u002Fblog\u002F2026\u002F08\u002F27\u002Fa-new-chapter-for-a2a-joining-the-agentic-ai-foundation\u002F\" target=\"_blank\" rel=\"noopener noreferrer\" class=\"editorjs-link-tool block border border-gray-200 dark:border-gray-700 rounded-lg p-4 transition text-gray-900 dark:text-gray-100 hover:border-primary-500 hover:bg-primary-50 dark:hover:bg-gray-900 hover:text-gray-900 dark:hover:text-gray-100\">\u003Cstrong class=\"block font-semibold\">A2A — Incorporación a la Agentic AI Foundation\u003C\u002Fstrong>\u003Cp class=\"text-sm text-gray-600 dark:text-gray-400\">Enfoque actual del proyecto de A2A como la capa horizontal de colaboración entre agentes junto con MCP como integración vertical de herramientas\u002Fdatos.\u003C\u002Fp>\u003C\u002Fa>\n\u003Ca href=\"https:\u002F\u002Fdevelopers.googleblog.com\u002Funder-the-hood-universal-commerce-protocol-ucp\u002F\" target=\"_blank\" rel=\"noopener noreferrer\" class=\"editorjs-link-tool block border border-gray-200 dark:border-gray-700 rounded-lg p-4 transition text-gray-900 dark:text-gray-100 hover:border-primary-500 hover:bg-primary-50 dark:hover:bg-gray-900 hover:text-gray-900 dark:hover:text-gray-100\">\u003Cstrong class=\"block font-semibold\">Google Developers — Bajo el capó: Universal Commerce Protocol\u003C\u002Fstrong>\u003Cp class=\"text-sm text-gray-600 dark:text-gray-400\">Descripción técnica general de UCP, sus primitivas de comercio y su capacidad de composición con APIs, A2A, MCP y AP2.\u003C\u002Fp>\u003C\u002Fa>\n\u003Ca href=\"https:\u002F\u002Fdevelopers.google.com\u002Fmerchant\u002Fucp\u002Fimplementation\u002F2026-04-08\u002Fpublish-profile\" target=\"_blank\" rel=\"noopener noreferrer\" class=\"editorjs-link-tool block border border-gray-200 dark:border-gray-700 rounded-lg p-4 transition text-gray-900 dark:text-gray-100 hover:border-primary-500 hover:bg-primary-50 dark:hover:bg-gray-900 hover:text-gray-900 dark:hover:text-gray-100\">\u003Cstrong class=\"block font-semibold\">Google Universal Commerce Protocol — Perfil UCP\u003C\u002Fstrong>\u003Cp class=\"text-sm text-gray-600 dark:text-gray-400\">Mecanismo actual de perfiles versionados para publicar servicios UCP y capacidades comerciales de comerciantes.\u003C\u002Fp>\u003C\u002Fa>\n\u003Ca href=\"https:\u002F\u002Fcloud.google.com\u002Fblog\u002Fproducts\u002Fai-machine-learning\u002Fannouncing-agents-to-payments-ap2-protocol\" target=\"_blank\" rel=\"noopener noreferrer\" class=\"editorjs-link-tool block border border-gray-200 dark:border-gray-700 rounded-lg p-4 transition text-gray-900 dark:text-gray-100 hover:border-primary-500 hover:bg-primary-50 dark:hover:bg-gray-900 hover:text-gray-900 dark:hover:text-gray-100\">\u003Cstrong class=\"block font-semibold\">Google Cloud — Agent Payments Protocol (AP2)\u003C\u002Fstrong>\u003Cp class=\"text-sm text-gray-600 dark:text-gray-400\">Anuncio y justificación de un protocolo abierto que cubre la autorización, autenticidad y rendición de cuentas en pagos gestionados por agentes.\u003C\u002Fp>\u003C\u002Fa>\n\u003Ca href=\"https:\u002F\u002Fdevelopers.googleblog.com\u002Fa2ui-v0-9-generative-ui\u002F\" target=\"_blank\" rel=\"noopener noreferrer\" class=\"editorjs-link-tool block border border-gray-200 dark:border-gray-700 rounded-lg p-4 transition text-gray-900 dark:text-gray-100 hover:border-primary-500 hover:bg-primary-50 dark:hover:bg-gray-900 hover:text-gray-900 dark:hover:text-gray-100\">\u003Cstrong class=\"block font-semibold\">Google Developers — A2UI v0.9\u003C\u002Fstrong>\u003Cp class=\"text-sm text-gray-600 dark:text-gray-400\">El modelo declarativo e independiente del framework de A2UI para interfaces portátiles impulsadas por agentes renderizadas por componentes nativos del host.\u003C\u002Fp>\u003C\u002Fa>\n\u003Ca href=\"https:\u002F\u002Fdevelopers.googleblog.com\u002Fa2ui-and-mcp-apps\u002F\" target=\"_blank\" rel=\"noopener noreferrer\" class=\"editorjs-link-tool block border border-gray-200 dark:border-gray-700 rounded-lg p-4 transition text-gray-900 dark:text-gray-100 hover:border-primary-500 hover:bg-primary-50 dark:hover:bg-gray-900 hover:text-gray-900 dark:hover:text-gray-100\">\u003Cstrong class=\"block font-semibold\">Google Developers — A2UI + MCP Apps\u003C\u002Fstrong>\u003Cp class=\"text-sm text-gray-600 dark:text-gray-400\">Cómo A2UI declarativo y las experiencias más ricas de MCP Apps pueden coexistir en lugar de tratarse como modelos de interfaz de usuario mutuamente excluyentes.\u003C\u002Fp>\u003C\u002Fa>",{"time":212,"blocks":213,"version":895},1790364209844,[214,222,228,236,243,250,255,260,265,270,306,313,318,323,328,333,345,351,356,361,366,371,376,386,391,423,428,433,438,443,448,453,463,468,473,478,483,488,515,520,525,530,535,540,550,555,560,589,594,618,623,628,633,638,643,676,681,686,691,700,705,710,715,720,725,730,735,740,745,750,780,785,808,813,823,832,841,850,859,868,877,886],{"id":215,"data":216,"type":220,"tunes":221},"_z4yTl5Fh-",{"title":217,"maxLevel":218,"minLevel":219},"Contenido",3,2,"tableOfContents",{},{"id":223,"data":224,"type":226,"tunes":227},"intro",{"text":225},"Los protocolos de agentes de IA se multiplican con rapidez: MCP, A2A, UCP, AP2, A2UI y estándares adyacentes aparecen cada vez más en los mismos diagramas de arquitectura. A menudo se describen como protocolos competidores. En la práctica, la mayoría resuelve diferentes problemas de interoperabilidad en límites distintos. La pregunta útil no es “¿Qué protocolo ganará?”, sino “¿Qué relación en el sistema necesita estandarizarse?”","paragraph",{},{"id":229,"data":230,"type":234,"tunes":235},"direct",{"body":231,"title":232,"variant":233},"\u003Cstrong>MCP, A2A, UCP, AP2 y A2UI son en su mayoría complementarios, no sustitutos.\u003C\u002Fstrong> MCP conecta una aplicación de IA con herramientas, datos y recursos. A2A conecta agentes independientes entre sí. UCP estandariza las interacciones de comercio entre interfaces de usuario para consumidores, empresas y ecosistemas de pago. AP2 añade autorización verificable e intención de pago a las transacciones lideradas por agentes. A2UI permite que un agente describa interfaces de usuario interactivas sin enviar código de aplicación arbitrario. Un sistema de producción puede utilizar legítimamente varios de ellos en un mismo flujo de trabajo.","Respuesta directa","info","callout",{},{"id":237,"data":238,"type":234,"tunes":242},"freshness",{"body":239,"title":240,"variant":241},"Este artículo refleja el panorama de protocolos a fecha de \u003Cstrong>25 de septiembre de 2026\u003C\u002Fstrong>. A2A ha alcanzado la versión 1.0, la línea actual de MCP TypeScript v2 implementa la especificación del 2026-07-28, UCP ya ha añadido versiones de protocolo más recientes de 2026 y A2UI continúa evolucionando. Siempre verifique la especificación actual antes de la implementación.","Estándares en rápida evolución","warning",{},{"id":244,"data":245,"type":234,"tunes":249},"model-note",{"body":246,"title":247,"variant":248},"La Pila de Responsabilidad de Protocolos y la Prueba de Selección de Protocolos a continuación son modelos prácticos de arquitectura propuestos aquí. No constituyen terminología oficial de los proyectos de los protocolos.","El modelo utilizado en este artículo","note",{},{"id":251,"data":252,"type":42,"tunes":254},"h-core",{"text":253,"level":219},"El error fundamental: comparar protocolos que operan en límites diferentes",{},{"id":256,"data":257,"type":226,"tunes":259},"p-core-1",{"text":258},"Un protocolo es útil porque dos sistemas implementados de forma independiente necesitan un contrato estable. El contrato solo tiene sentido si el límite está claro. Un agente que se comunica con una base de datos tiene un problema de interoperabilidad diferente al de un agente que delega trabajo en otro, un comprador que autoriza una compra o un agente remoto que solicita a una aplicación nativa que renderice un formulario.",{},{"id":261,"data":262,"type":226,"tunes":264},"p-core-2",{"text":263},"La guía de desarrolladores de Google de 2026 presenta explícitamente MCP, A2A, UCP, AP2, A2UI y los protocolos de interfaz de usuario relacionados como una pila de estándares complementarios. Un mismo flujo de trabajo de ejemplo puede utilizar varios de ellos de forma conjunta: herramientas para el inventario, agentes remotos para proveedores, comercio para pedidos, autorización de pagos para gastos y protocolos de interfaz de usuario para la interacción.",{},{"id":266,"data":267,"type":42,"tunes":269},"h-stack",{"text":268,"level":219},"La Pila de Responsabilidad de Protocolos",{},{"id":271,"data":272,"type":304,"tunes":305},"stack-table",{"content":273,"stretched":43,"withHeadings":14},[274,279,284,289,294,299],[275,276,277,278],"Protocolo","¿Qué relación estandariza?","Abstracción principal","No diseñado principalmente para",[280,281,282,283],"MCP","Aplicación de IA ↔ herramientas, recursos y datos","Herramientas, recursos, prompts e intercambio de capacidades entre host y servidor","Colaboración independiente entre agentes o semántica comercial",[285,286,287,288],"A2A","Agente ↔ agente independiente","Descubrimiento de agentes, mensajes, tareas, artefactos y colaboración de larga duración","Integración directa de bases de datos\u002Fherramientas",[290,291,292,293],"UCP","Superficie de consumidor\u002Fagente ↔ sistema comercial de comercios","Capacidades de producto\u002Fcarrito\u002Fpago (checkout)\u002Fcumplimiento\u002Fpedidos","Comunicación de propósito general entre agentes",[295,296,297,298],"AP2","Intención del usuario\u002Fagente ↔ autorización de pago","Mandatos, restricciones de aprobación y autoridad auditable de pagos liderados por agentes","Descubrimiento de productos o transporte genérico de pasarelas de pago",[300,301,302,303],"A2UI","Agente ↔ host de interfaz de usuario","Intención declarativa de UI renderizada mediante componentes nativos confiables","Código frontend remoto arbitrario o delegación de tareas entre agentes","table",{},{"id":307,"data":308,"type":234,"tunes":312},"all-five",{"body":309,"title":310,"variant":311},"Un agente puede usar \u003Cstrong>MCP\u003C\u002Fstrong> para inspeccionar el inventario, \u003Cstrong>A2A\u003C\u002Fstrong> para consultar la disponibilidad a un agente proveedor, \u003Cstrong>UCP\u003C\u002Fstrong> para estructurar una transacción comercial, \u003Cstrong>AP2\u003C\u002Fstrong> para certificar la autorización de gasto y \u003Cstrong>A2UI\u003C\u002Fstrong> para renderizar una interfaz de aprobación nativa para el usuario.","Un único flujo de trabajo puede utilizar los cinco","success",{},{"id":314,"data":315,"type":42,"tunes":317},"h-mcp",{"text":316,"level":219},"1. MCP: conectar el agente a las capacidades",{},{"id":319,"data":320,"type":226,"tunes":322},"p-mcp-1",{"text":321},"El Model Context Protocol es un estándar abierto para conectar aplicaciones de IA a sistemas externos donde residen herramientas, datos y recursos reutilizables. Un servidor expone capacidades; un host de MCP se conecta a ese servidor y pone dichas capacidades a disposición del modelo o aplicación.",{},{"id":324,"data":325,"type":226,"tunes":327},"p-mcp-2",{"text":326},"La documentación actual de MCP TypeScript v2 describe el protocolo exactamente en esos términos: los servidores exponen herramientas, recursos y prompts, mientras que los hosts, como entornos de desarrollo o aplicaciones personalizadas, se conectan a ellos. Esto convierte a MCP principalmente en un protocolo de integración de capacidades.",{},{"id":329,"data":330,"type":42,"tunes":332},"h-mcp-use",{"text":331,"level":218},"Utilice MCP cuando",{},{"id":334,"data":335,"type":343,"tunes":344},"mcp-list",{"meta":336,"items":337,"style":342},{},[338,339,340,341],"Una aplicación de IA necesite acceso estandarizado a herramientas o API.","Desea que un único servidor de capacidades funcione con múltiples hosts de IA compatibles.","Necesite acceso estructurado a datos o recursos sin codificar de forma fija cada integración en cada agente.","El sistema externo sea un proveedor de capacidades, no un agente autónomo equivalente.","unordered","list",{},{"id":346,"data":347,"type":234,"tunes":350},"mcp-warning",{"body":348,"title":349,"variant":241},"Un servidor MCP puede exponer funciones avanzadas, pero eso no lo convierte en un agente independiente con su propio ciclo de vida de tareas, semántica de descubrimiento y razonamiento interno opaco. La invocación de herramientas y la colaboración entre agentes son contratos distintos.","MCP no equivale automáticamente a A2A",{},{"id":352,"data":353,"type":42,"tunes":355},"h-a2a",{"text":354,"level":219},"2. A2A: conectar agentes independientes",{},{"id":357,"data":358,"type":226,"tunes":360},"p-a2a-1",{"text":359},"Agent2Agent (A2A) está diseñado para la comunicación entre sistemas de agentes independientes y potencialmente opacos. Su especificación actual v1.0 se centra en el descubrimiento de capacidades, mensajería, tareas, artefactos, contenido multimodal y colaboración de larga duración sin requerir que un agente exponga sus herramientas internas, memoria o implementación a otro.",{},{"id":362,"data":363,"type":226,"tunes":365},"p-a2a-2",{"text":364},"Esa opacidad es el límite importante. El agente invocador no necesita saber si el agente remoto utiliza internamente MCP, herramientas personalizadas, un planificador propietario, otro proveedor de modelos o escalamiento humano. Necesita un contrato para descubrir capacidades y delegar trabajo.",{},{"id":367,"data":368,"type":226,"tunes":370},"p-a2a-3",{"text":369},"A2A v1.0 también estandariza la negociación de versiones y admite múltiples enlaces alrededor de un modelo de datos común. Su mecanismo publicado de Agent Card ofrece a los clientes un punto de descubrimiento estándar para las capacidades, protocolos admitidos, requisitos de autenticación y habilidades de un agente.",{},{"id":372,"data":373,"type":42,"tunes":375},"h-a2a-use",{"text":374,"level":218},"Use A2A cuando",{},{"id":377,"data":378,"type":343,"tunes":385},"a2a-list",{"meta":379,"items":380,"style":342},{},[381,382,383,384],"Un agente autónomo necesita delegar trabajo en otro agente autónomo.","El sistema remoto debe permanecer opaco detrás de un contrato de capacidades.","Las tareas pueden ser de larga duración, asíncronas o requerir interacción humana en el bucle (human-in-the-loop).","Los agentes están construidos con diferentes frameworks, lenguajes, proveedores o pertenencia organizativa.",{},{"id":387,"data":388,"type":42,"tunes":390},"h-mcp-a2a",{"text":389,"level":219},"MCP vs. A2A: integración vertical frente a colaboración horizontal",{},{"id":392,"data":393,"type":421,"tunes":422},"mcp-a2a-comparison",{"rows":394,"title":412,"layout":304,"columns":413},[395,400,404,408],{"id":396,"label":397,"values":398},"relationship","Relación",[399,399,399],"",{"id":401,"label":402,"values":403},"abstraction","Abstracción",[399,399,399],{"id":405,"label":406,"values":407},"opacity","Opacidad interna",[399,399,399],{"id":409,"label":410,"values":411},"long","Trabajo de larga duración",[399,399,399],"MCP y A2A resuelven diferentes problemas de interoperabilidad",[414,417,419],{"id":415,"label":416},"dimension","Dimensión",{"id":418,"label":280},"mcp",{"id":420,"label":285},"a2a","comparison",{},{"id":424,"data":425,"type":226,"tunes":427},"p-mcp-a2a-1",{"text":426},"El propio proyecto A2A describe ahora la distinción como horizontal frente a vertical: MCP conecta agentes con herramientas y bases de datos internas, mientras que A2A permite la colaboración peer-to-peer entre sistemas de agentes.",{},{"id":429,"data":430,"type":42,"tunes":432},"h-ucp",{"text":431,"level":219},"3. UCP: estandarizar el comercio agéntico",{},{"id":434,"data":435,"type":226,"tunes":437},"p-ucp-1",{"text":436},"El Universal Commerce Protocol no es un protocolo de agentes genérico. Estandariza los recorridos comerciales entre interfaces de consumidores, comerciantes y proveedores de pagos. La implementación de Google ya admite capacidades como la creación de carritos, el checkout, el cumplimiento de pedidos y el ciclo de vida de los pedidos a través de perfiles y API versionadas.",{},{"id":439,"data":440,"type":226,"tunes":442},"p-ucp-2",{"text":441},"Un comerciante puede publicar un perfil de UCP en \u002F.well-known\u002Fucp describiendo servicios, versiones de protocolo y capacidades. Ese patrón de descubrimiento es importante porque una interfaz agéntica no debería necesitar un contrato de checkout a medida para cada comerciante.",{},{"id":444,"data":445,"type":226,"tunes":447},"p-ucp-3",{"text":446},"UCP también es intencionalmente componible. La descripción técnica general de Google indica que puede integrarse a través de API, A2A y MCP, y es compatible con AP2 para la autorización de pagos agénticos.",{},{"id":449,"data":450,"type":42,"tunes":452},"h-ucp-use",{"text":451,"level":218},"Use UCP cuando",{},{"id":454,"data":455,"type":343,"tunes":462},"ucp-list",{"meta":456,"items":457,"style":342},{},[458,459,460,461],"El flujo de trabajo involucra productos de comerciantes, carritos, checkout, cumplimiento o ciclo de vida del pedido.","Está construyendo una interfaz para comerciantes que debe funcionar con experiencias de compra agénticas.","La integración necesita semántica específica de comercio en lugar de llamadas a herramientas genéricas.","Desea un contrato de comercio interoperable que pueda coexistir con MCP, A2A y protocolos de pago.",{},{"id":464,"data":465,"type":42,"tunes":467},"h-ap2",{"text":466,"level":219},"4. AP2: demostrar que el agente tenía autorización para gastar",{},{"id":469,"data":470,"type":226,"tunes":472},"p-ap2-1",{"text":471},"El comercio agéntico introduce un problema que los flujos de checkout ordinarios no tenían que resolver de la misma manera: un agente puede realizar transacciones cuando el humano no está haciendo clic en el botón final en tiempo real. El Agent Payments Protocol (AP2) aborda la autorización, la autenticidad y la rendición de cuentas para los pagos dirigidos por agentes.",{},{"id":474,"data":475,"type":226,"tunes":477},"p-ap2-2",{"text":476},"La guía de protocolos de Google de 2026 describe AP2 a través de mandatos tipados que capturan la intención del usuario, las restricciones de gasto y la transacción específica que se autoriza. AP2 puede funcionar como una extensión junto a UCP: UCP describe la transacción comercial, mientras que AP2 proporciona pruebas de que el agente tenía autorización para realizar el pago.",{},{"id":479,"data":480,"type":226,"tunes":482},"p-ap2-3",{"text":481},"Esta distinción es importante. Un protocolo de checkout puede indicarle a un comerciante qué debe comprarse. Por sí solo no demuestra quién autorizó al agente a gastar, bajo qué límite, para qué comerciante, durante cuánto tiempo o si el carrito final se mantuvo dentro de esa autorización.",{},{"id":484,"data":485,"type":42,"tunes":487},"h-ucp-ap2",{"text":486,"level":219},"UCP vs. AP2: semántica de transacciones frente a autoridad",{},{"id":489,"data":490,"type":304,"tunes":514},"ucp-ap2-table",{"content":491,"stretched":43,"withHeadings":14},[492,494,498,502,506,510],[493,290,295],"Pregunta",[495,496,497],"¿Qué se está comprando?","Artículos de comercio, carrito, semántica de checkout y cumplimiento","Hace referencia al contexto de la transacción autorizada",[499,500,501],"¿Quién puede autorizarlo?","No es la responsabilidad principal del protocolo","Modelo explícito de autoridad y mandatos de agente\u002Fusuario",[503,504,505],"¿Qué restricciones de gasto aplican?","El flujo de comercio puede contener totales y datos de checkout","Límites de autorización y restricciones de intención",[507,508,509],"¿Cómo se audita la transacción?","Ciclo de vida del pedido y del comercio","Rastro de autorización criptográfico \u002F verificable mediante mandatos y recibos",[511,512,513],"¿Pueden trabajar juntos?","Sí","Sí: AP2 puede ampliar los flujos de comercio agéntico",{},{"id":516,"data":517,"type":42,"tunes":519},"h-a2ui",{"text":518,"level":219},"5. A2UI: permita que los agentes describan interfaces sin adueñarse de su frontend",{},{"id":521,"data":522,"type":226,"tunes":524},"p-a2ui-1",{"text":523},"Agent-to-User Interface (A2UI) aborda otro límite: cómo un agente remoto o local comunica una interfaz interactiva enriquecida a una aplicación anfitriona. En lugar de enviar HTML, CSS y JavaScript arbitrarios, A2UI utiliza datos declarativos que el anfitrión renderiza a través de su propio catálogo de componentes de confianza.",{},{"id":526,"data":527,"type":226,"tunes":529},"p-a2ui-2",{"text":528},"Esto preserva el sistema de diseño y el modelo de seguridad de la aplicación anfitriona, al tiempo que permite que un agente solicite interfaces dinámicas. A2UI v0.9 enfatiza específicamente la intención de interfaz de usuario independiente del framework y las actualizaciones en streaming a través de la web, móviles y otros clientes.",{},{"id":531,"data":532,"type":226,"tunes":534},"p-a2ui-3",{"text":533},"El trabajo posterior de Google sobre A2UI + MCP Apps también demuestra que estos modelos de interfaz de usuario no son necesariamente mutuamente excluyentes. La interfaz de usuario nativa declarativa y las experiencias más ricas de aplicaciones integradas pueden coexistir según la tarea.",{},{"id":536,"data":537,"type":42,"tunes":539},"h-a2ui-use",{"text":538,"level":218},"Utilice A2UI cuando",{},{"id":541,"data":542,"type":343,"tunes":549},"a2ui-list",{"meta":543,"items":544,"style":342},{},[545,546,547,548],"Un agente remoto necesita solicitar formularios, tarjetas, controles u otra interfaz interactiva.","El anfitrión debe preservar sus componentes nativos, estilo y límite de seguridad.","No desea que los agentes remotos envíen código frontend ejecutable arbitrario.","La misma intención de interfaz definida por el agente debe funcionar en diferentes frameworks de cliente.",{},{"id":551,"data":552,"type":42,"tunes":554},"h-selection",{"text":553,"level":219},"La prueba de selección de protocolos",{},{"id":556,"data":557,"type":226,"tunes":559},"p-selection-intro",{"text":558},"No empiece por el acrónimo. Empiece por la relación que necesita interoperabilidad.",{},{"id":561,"data":562,"type":587,"tunes":588},"selection-flow",{"steps":563,"title":585,"orientation":586},[564,567,570,573,576,579,582],{"label":565,"description":566},"1. Identifique las dos partes independientes","¿Se trata de IA a herramienta, agente a agente, agente a comerciante, agente a autoridad de pago o agente a interfaz de usuario?",{"label":568,"description":569},"2. Identifique el objeto compartido","¿El contrato trata sobre una llamada a herramienta, tarea, carrito, mandato de pago, artefacto o descripción de interfaz de usuario?",{"label":571,"description":572},"3. Compruebe si ya existe un protocolo de dominio","Prefiera la semántica de comercio o pago cuando el problema sea de comercio o autorización, en lugar de codificar todo como herramientas genéricas.",{"label":574,"description":575},"4. Mantenga los aspectos internos locales como locales","No exponga un agente completo como herramientas MCP si la parte remota solo necesita una capacidad A2A, y no haga que un agente remoto sea responsable del tiempo de ejecución de su interfaz de usuario.",{"label":577,"description":578},"5. Componga protocolos cuando el flujo de trabajo cruce límites","Un flujo de trabajo puede cruzar legítimamente contratos de herramientas, agentes, comercio, pagos e interfaz de usuario.",{"label":580,"description":581},"6. Controle las versiones de cada contrato de forma independiente","Las versiones de los protocolos evolucionan a diferentes velocidades; no vincule cada integración a una única versión monolítica de la aplicación.",{"label":583,"description":584},"7. Preserve la autorización en cada límite","La interoperabilidad no reemplaza los permisos de producto, la autorización de herramientas, la autoridad de pago ni la política de acceso a datos.","Elija el protocolo según el límite","auto","processFlow",{},{"id":590,"data":591,"type":42,"tunes":593},"h-workflow",{"text":592,"level":219},"Un flujo de trabajo multiprotocolo realista",{},{"id":595,"data":596,"type":587,"tunes":617},"workflow-flow",{"steps":597,"title":616,"orientation":586},[598,601,604,607,610,613],{"label":599,"description":600},"1. Inspeccionar el stock interno con MCP","El agente de compras llama a las capacidades de inventario y previsión expuestas por los servidores MCP internos.",{"label":602,"description":603},"2. Descubrir un agente proveedor con A2A","El agente lee la Agent Card del proveedor y delega una tarea de disponibilidad y tiempo de entrega.",{"label":605,"description":606},"3. Negociar el objeto comercial con UCP","La superficie del proveedor o comerciante devuelve información estructurada sobre el carrito, el checkout y el cumplimiento.",{"label":608,"description":609},"4. Verificar la autoridad de gasto con AP2","La compra se compara con el mandato firmado del usuario o de la organización, las restricciones del comerciante y los límites de gasto.",{"label":611,"description":612},"5. Solicitar aprobación a través de A2UI","Si se requiere aprobación humana, el agente envía una intención declarativa de interfaz de usuario y el anfitrión renderiza la experiencia de aprobación utilizando componentes nativos confiables.",{"label":614,"description":615},"6. Completar y auditar","El estado del comercio, la autorización de pago, la evidencia de la tarea del agente y los registros de auditoría de la aplicación siguen siendo rastreables a través de sus respectivos límites.","Ejemplo: un flujo de trabajo de aprovisionamiento autónomo",{},{"id":619,"data":620,"type":42,"tunes":622},"h-universal",{"text":621,"level":219},"Por qué es poco probable que un protocolo universal de agentes los reemplace a todos",{},{"id":624,"data":625,"type":226,"tunes":627},"p-universal-1",{"text":626},"Un protocolo universal suena más sencillo hasta que debe codificar la semántica de cada dominio. El descubrimiento de herramientas, la colaboración prolongada entre agentes, el checkout, la autorización de pagos y la interfaz de usuario nativa tienen diferentes requisitos de ciclo de vida, seguridad y corrección.",{},{"id":629,"data":630,"type":226,"tunes":632},"p-universal-2",{"text":631},"La propia web evolucionó mediante protocolos por capas en lugar de un único formato de mensaje para cada problema. El stack agéntico emergente parece avanzar en la misma dirección: primitivas horizontales comunes, contratos de dominio especializados y descubrimiento\u002Fversionado explícitos.",{},{"id":634,"data":635,"type":226,"tunes":637},"p-universal-3",{"text":636},"Por lo tanto, el desafío arquitectónico pasa de «¿qué protocolo gana?» a con qué claridad se componen los protocolos sin duplicar la identidad, la autorización, el estado y la semántica de auditoría.",{},{"id":639,"data":640,"type":42,"tunes":642},"h-failures",{"text":641,"level":219},"La composición de protocolos genera nuevos modos de fallo",{},{"id":644,"data":645,"type":304,"tunes":675},"failure-table",{"content":646,"stretched":43,"withHeadings":14},[647,651,655,659,663,667,671],[648,649,650],"Modo de fallo","Qué sucede","Control de arquitectura",[652,653,654],"Fuga de autoridad","Una capacidad válida de herramienta o agente se trata como permiso para realizar una acción de negocio","Mantener la autorización del producto independiente del descubrimiento de capacidades del protocolo",[656,657,658],"Discrepancia de identidad","La identidad del host de MCP, la identidad del agente A2A y la identidad de comercio\u002Fpago hacen referencia a diferentes entidades principales","Definir un mapeo explícito de entidades principales a través de los límites",[660,661,662],"Desfase de versiones","Un protocolo se actualiza mientras que los adaptadores dependientes asumen una semántica anterior","Negociar y fijar versiones de protocolo de forma independiente",[664,665,666],"Duplicación de estado","El mismo estado de carrito, tarea o aprobación se copia en varias capas de protocolo","Definir un único propietario con autoridad por objeto de dominio",[668,669,670],"Fragmentación de auditoría","Las trazas de herramientas, las tareas de agentes y la evidencia de checkout y pagos no se pueden vincular","Transmitir identificadores de correlación e identificadores de dominio estables a través de los límites del protocolo",[672,673,674],"Tunelización semántica","Todo se fuerza a través de un protocolo genérico como JSON opaco","Utilizar protocolos de dominio donde su semántica mejore materialmente la corrección",{},{"id":677,"data":678,"type":42,"tunes":680},"h-app-arch",{"text":679,"level":219},"La elección del protocolo no reemplaza la arquitectura de la aplicación",{},{"id":682,"data":683,"type":226,"tunes":685},"p-app-1",{"text":684},"Los estándares abiertos reducen el acoplamiento de integración, pero no deciden su modelo de dominio, política de autorización, fuente de verdad, estrategia de reintentos o criterios de aceptación. Una herramienta MCP aún puede exponer la capacidad incorrecta. Un agente A2A aún puede devolver un artefacto erróneo. Un checkout de UCP aún puede contener datos desactualizados del comerciante. La lógica de la aplicación aún puede aplicar incorrectamente un mandato AP2.",{},{"id":687,"data":688,"type":226,"tunes":690},"p-app-2",{"text":689},"Trate los protocolos como contratos entre componentes que evolucionan de forma independiente. Mantenga la verdad del dominio y la política consecuente en la capa de aplicación a la que pertenecen, y luego use protocolos para hacer que los límites sean interoperables.",{},{"id":692,"data":693,"type":698,"tunes":699},"ref-harness",{"url":694,"title":695,"excerpt":696,"ctaLabel":697},"https:\u002F\u002Fstajic.de\u002Fes\u002Fblog\u002Fmanaged-agent-harness-vs-self-hosted-agent-loop-what-you-gain-what-you-lose","Harness de agentes gestionado vs. bucle de agentes autohospedado: lo que gana, lo que pierde","Los límites de protocolo resuelven la interoperabilidad. Los límites de tiempo de ejecución resuelven quién opera el harness, el entorno de ejecución y el plano de control de la aplicación.","Leer la guía de arquitectura de tiempo de ejecución","referralArticle",{},{"id":701,"data":702,"type":42,"tunes":704},"h-change",{"text":703,"level":219},"¿Qué cambiaría esta respuesta?",{},{"id":706,"data":707,"type":226,"tunes":709},"p-change-1",{"text":708},"La pila tecnológica cambia si los protocolos convergen, si un estándar absorbe formalmente a otro o si los proveedores estandarizan una capa compartida de identidad y autorización a través de múltiples límites. UCP ya demuestra la composición al admitir APIs, A2A y MCP, y al integrarse con AP2 en lugar de reemplazarlos.",{},{"id":711,"data":712,"type":226,"tunes":714},"p-change-2",{"text":713},"La respuesta también varía según el alcance de la aplicación. Un agente interno pequeño puede necesitar únicamente MCP. Un flujo de trabajo entre múltiples empresas puede necesitar A2A. Un comerciante puede necesitar UCP sin A2UI. Un agente de compras delegado puede necesitarlos todos. Utilice el conjunto de protocolos más pequeño que represente los límites reales sin aplanar la semántica de dominio.",{},{"id":716,"data":717,"type":42,"tunes":719},"h-limitations",{"text":718,"level":219},"Limitaciones",{},{"id":721,"data":722,"type":226,"tunes":724},"p-limit-1",{"text":723},"Los protocolos analizados aquí se encuentran en diferentes niveles de madurez y tienen distintos modelos de gobernanza. A2A ha alcanzado una especificación v1.0 estable, mientras que otros estándares continúan evolucionando rápidamente. La adopción en el ecosistema también es desigual entre proveedores y frameworks.",{},{"id":726,"data":727,"type":226,"tunes":729},"p-limit-2",{"text":728},"Este artículo se enfoca en la responsabilidad arquitectónica más que en la exhaustividad de la implementación. Los métodos de autenticación específicos, los enlaces de transporte, los esquemas y los mecanismos de extensión deben tomarse de la especificación vigente de cada protocolo.",{},{"id":731,"data":732,"type":42,"tunes":734},"h-conclusion",{"text":733,"level":219},"Conclusión",{},{"id":736,"data":737,"type":226,"tunes":739},"p-conclusion-1",{"text":738},"MCP, A2A, UCP, AP2 y A2UI cobran más sentido cuando se consideran protocolos para diferentes relaciones, y no cinco intentos competitivos de estandarizar los “agentes”.",{},{"id":741,"data":742,"type":226,"tunes":744},"p-conclusion-2",{"text":743},"MCP expone capacidades. A2A coordina agentes independientes. UCP otorga al comercio su propio contrato legible por máquinas. AP2 añade autoridad de pago verificable. A2UI proporciona a los agentes una vía declarativa y segura hacia las interfaces de usuario. Por lo tanto, la web agéntica emergente no está reemplazando los protocolos con IA; está creando una nueva pila de protocolos en torno a la IA.",{},{"id":746,"data":747,"type":42,"tunes":749},"h-faq",{"text":748,"level":219},"Preguntas frecuentes",{},{"id":751,"data":752,"type":751,"tunes":779},"faq",{"items":753,"title":778},[754,758,762,766,770,774],{"id":755,"answer":756,"question":757},"faq1","No. MCP estandariza principalmente la forma en que las aplicaciones de IA acceden a herramientas, recursos y datos. A2A estandariza la colaboración entre sistemas de agentes independientes. Un agente remoto puede usar MCP internamente mientras expone una interfaz A2A.","¿Es A2A un reemplazo de MCP?",{"id":759,"answer":760,"question":761},"faq2","Generalmente no. UCP proporciona semántica específica de comercio, como carrito, checkout y cumplimiento. MCP aún puede exponer herramientas o datos del comerciante, y UCP está diseñado para coexistir con MCP y A2A.","¿Es UCP un reemplazo de MCP en agentes de compras?",{"id":763,"answer":764,"question":765},"faq3","UCP estandariza las interacciones comerciales y el ciclo de vida de las transacciones. AP2 se enfoca en demostrar que un agente tenía autorización para realizar un pago bajo restricciones definidas del usuario o de la organización.","¿Cuál es la diferencia entre UCP y AP2?",{"id":767,"answer":768,"question":769},"faq4","A2UI permite que los agentes envíen intenciones de interfaz de usuario declarativas a una aplicación host, la cual renderiza la experiencia mediante componentes nativos confiables en lugar de ejecutar código frontend remoto arbitrario.","¿Qué problema resuelve A2UI?",{"id":771,"answer":772,"question":773},"faq5","Sí. Un flujo de trabajo puede utilizar MCP para herramientas internas, A2A para delegación a agentes remotos, UCP para comercio, AP2 para autorización de pagos y A2UI para interacción humana.","¿Puede una aplicación de agentes utilizar todos estos protocolos?",{"id":775,"answer":776,"question":777},"faq6","Comience a partir del límite de interoperabilidad. Si el problema es el acceso a herramientas, evalúe MCP. Si es la colaboración entre agentes independientes, evalúe A2A. Si es de comercio, UCP. Si es la autoridad de pago delegada, AP2. Si es una interfaz de usuario portátil impulsada por agentes, A2UI.","¿Qué protocolo debería implementar primero?","MCP, A2A, UCP, AP2 y A2UI",{},{"id":781,"data":782,"type":42,"tunes":784},"h-glossary",{"text":783,"level":219},"Glosario",{},{"id":786,"data":787,"type":786,"tunes":807},"glossary",{"title":788,"entries":789},"Términos clave sobre protocolos de agentes",[790,792,794,797,800,803],{"term":280,"anchor":418,"definition":791},"Model Context Protocol, un estándar abierto para exponer herramientas, recursos y prompts desde sistemas externos hacia hosts de IA compatibles.",{"term":285,"anchor":420,"definition":793},"Agent2Agent Protocol, un estándar abierto para descubrir y colaborar con sistemas de agentes independientes a través de mensajes, tareas y artefactos.",{"term":290,"anchor":795,"definition":796},"ucp","Universal Commerce Protocol, un estándar abierto para recorridos de comercio agéntico interoperables entre interfaces de consumo, empresas y proveedores de pagos.",{"term":295,"anchor":798,"definition":799},"ap2","Agent Payments Protocol, un estándar abierto para representar y verificar la autoridad, la intención y la rendición de cuentas en pagos gestionados por agentes.",{"term":300,"anchor":801,"definition":802},"a2ui","Agent-to-User Interface, un protocolo declarativo para permitir que los agentes soliciten una interfaz de usuario que se renderiza utilizando los componentes confiables de la aplicación host.",{"term":804,"anchor":805,"definition":806},"Composición de protocolos","protocol-composition","El uso de múltiples protocolos en un solo flujo de trabajo, donde cada uno es responsable de un límite de interoperabilidad específico en lugar de forzar toda la semántica a través de un único contrato.",{},{"id":809,"data":810,"type":42,"tunes":812},"h-sources",{"text":811,"level":219},"Fuentes principales y lecturas adicionales",{},{"id":814,"data":815,"type":821,"tunes":822},"src-google-guide",{"link":816,"meta":817},"https:\u002F\u002Fdevelopers.googleblog.com\u002Fdevelopers-guide-to-ai-agent-protocols\u002F",{"image":818,"title":819,"description":820},{"url":399},"Google Developers — Guía para desarrolladores sobre protocolos de agentes de IA","Una descripción práctica que muestra cómo MCP, A2A, UCP, AP2, A2UI y protocolos relacionados funcionan juntos en un flujo de trabajo de agente de varios pasos.","linkTool",{},{"id":824,"data":825,"type":821,"tunes":831},"src-mcp",{"link":826,"meta":827},"https:\u002F\u002Fts.sdk.modelcontextprotocol.io\u002Fv2\u002F",{"image":828,"title":829,"description":830},{"url":399},"Model Context Protocol — TypeScript SDK v2","Documentación actual y estable del SDK que implementa la especificación MCP del 28-07-2026 y define herramientas, recursos, prompts e integración de host\u002Fservidor.",{},{"id":833,"data":834,"type":821,"tunes":840},"src-a2a-spec",{"link":835,"meta":836},"https:\u002F\u002Fa2a-protocol.org\u002Fdev\u002Fspecification\u002F",{"image":837,"title":838,"description":839},{"url":399},"Protocolo A2A — Especificación v1.0","Especificación actual del protocolo A2A que abarca Agent Cards, mensajes, tareas, artefactos, enlaces y negociación de versiones.",{},{"id":842,"data":843,"type":821,"tunes":849},"src-a2a-aaif",{"link":844,"meta":845},"https:\u002F\u002Fa2a-protocol.org\u002Flatest\u002Fblog\u002F2026\u002F08\u002F27\u002Fa-new-chapter-for-a2a-joining-the-agentic-ai-foundation\u002F",{"image":846,"title":847,"description":848},{"url":399},"A2A — Incorporación a la Agentic AI Foundation","Enfoque actual del proyecto de A2A como la capa horizontal de colaboración entre agentes junto con MCP como integración vertical de herramientas\u002Fdatos.",{},{"id":851,"data":852,"type":821,"tunes":858},"src-ucp",{"link":853,"meta":854},"https:\u002F\u002Fdevelopers.googleblog.com\u002Funder-the-hood-universal-commerce-protocol-ucp\u002F",{"image":855,"title":856,"description":857},{"url":399},"Google Developers — Bajo el capó: Universal Commerce Protocol","Descripción técnica general de UCP, sus primitivas de comercio y su capacidad de composición con APIs, A2A, MCP y AP2.",{},{"id":860,"data":861,"type":821,"tunes":867},"src-ucp-profile",{"link":862,"meta":863},"https:\u002F\u002Fdevelopers.google.com\u002Fmerchant\u002Fucp\u002Fimplementation\u002F2026-04-08\u002Fpublish-profile",{"image":864,"title":865,"description":866},{"url":399},"Google Universal Commerce Protocol — Perfil UCP","Mecanismo actual de perfiles versionados para publicar servicios UCP y capacidades comerciales de comerciantes.",{},{"id":869,"data":870,"type":821,"tunes":876},"src-ap2",{"link":871,"meta":872},"https:\u002F\u002Fcloud.google.com\u002Fblog\u002Fproducts\u002Fai-machine-learning\u002Fannouncing-agents-to-payments-ap2-protocol",{"image":873,"title":874,"description":875},{"url":399},"Google Cloud — Agent Payments Protocol (AP2)","Anuncio y justificación de un protocolo abierto que cubre la autorización, autenticidad y rendición de cuentas en pagos gestionados por agentes.",{},{"id":878,"data":879,"type":821,"tunes":885},"src-a2ui",{"link":880,"meta":881},"https:\u002F\u002Fdevelopers.googleblog.com\u002Fa2ui-v0-9-generative-ui\u002F",{"image":882,"title":883,"description":884},{"url":399},"Google Developers — A2UI v0.9","El modelo declarativo e independiente del framework de A2UI para interfaces portátiles impulsadas por agentes renderizadas por componentes nativos del host.",{},{"id":887,"data":888,"type":821,"tunes":894},"src-a2ui-mcp",{"link":889,"meta":890},"https:\u002F\u002Fdevelopers.googleblog.com\u002Fa2ui-and-mcp-apps\u002F",{"image":891,"title":892,"description":893},{"url":399},"Google Developers — A2UI + MCP Apps","Cómo A2UI declarativo y las experiencias más ricas de MCP Apps pueden coexistir en lugar de tratarse como modelos de interfaz de usuario mutuamente excluyentes.",{},"2.31","MCP, A2A, UCP, AP2 y A2UI a menudo se presentan como estándares de agentes competidores. En su mayoría, resuelven diferentes problemas de interoperabilidad. Esta guía mapea cada protocolo con el límite que realmente estandariza—y muestra cómo pueden funcionar juntos en un sistema de producción.","\u002Fuploads\u002F2026\u002F09\u002Fmcp-vs-a2a-vs-ucp-vs-ap2-vs-a2ui-the-agent-protocol-stack-explained-1790352625869-2ezle0.webp","mcp-vs-a2a-vs-ucp-vs-ap2-vs-a2ui-the-agent-protocol-stack-explained-1790352625869-2ezle0","PUBLISHED","2026-09-25T12:09:00.000Z","2026-09-25T16:09:26.909Z","2026-09-25T19:27:42.182Z",{"en":904,"de":905,"sr":906,"es":907,"fr":908,"it":909,"ru":910,"zh":911},"\u002Fblog\u002Fmcp-vs-a2a-vs-ucp-vs-ap2-vs-a2ui-the-agent-protocol-stack-explained","\u002Fde\u002Fblog\u002Fmcp-vs-a2a-vs-ucp-vs-ap2-vs-a2ui-the-agent-protocol-stack-explained","\u002Fsr\u002Fblog\u002Fmcp-vs-a2a-vs-ucp-vs-ap2-vs-a2ui-the-agent-protocol-stack-explained","\u002Fes\u002Fblog\u002Fmcp-vs-a2a-vs-ucp-vs-ap2-vs-a2ui-the-agent-protocol-stack-explained","\u002Ffr\u002Fblog\u002Fmcp-vs-a2a-vs-ucp-vs-ap2-vs-a2ui-the-agent-protocol-stack-explained","\u002Fit\u002Fblog\u002Fmcp-vs-a2a-vs-ucp-vs-ap2-vs-a2ui-the-agent-protocol-stack-explained","\u002Fru\u002Fblog\u002Fmcp-vs-a2a-vs-ucp-vs-ap2-vs-a2ui-the-agent-protocol-stack-explained","\u002Fzh\u002Fblog\u002Fmcp-vs-a2a-vs-ucp-vs-ap2-vs-a2ui-the-agent-protocol-stack-explained",[913,917,921],{"id":914,"name":915,"slug":916},57,"Límites de datos","data-boundaries",{"id":918,"name":919,"slug":920},84,"Política y límites de datos","policy-and-data",{"id":922,"name":923,"slug":924},46,"Resumen","overview",{"id":926,"login":927,"email":928,"displayName":929},"20","rooth8233","aleksandar@stajic.de","Aleksandar Stajić",[931,1486],{"lang":932,"title":933,"content":934,"contentJson":935,"excerpt":1485},"en","MCP vs A2A vs UCP vs AP2 vs A2UI: The Agent Protocol Stack Explained","{\"time\":1790352626854,\"blocks\":[{\"id\":\"_z4yTl5Fh-\",\"type\":\"tableOfContents\",\"data\":{\"title\":\"Contents\",\"minLevel\":2,\"maxLevel\":3},\"tunes\":{}},{\"id\":\"intro\",\"type\":\"paragraph\",\"data\":{\"text\":\"AI agent protocols are multiplying quickly: MCP, A2A, UCP, AP2, A2UI and adjacent standards increasingly appear in the same architecture diagrams. They are often described as competing protocols. In practice, most of them solve different interoperability problems at different boundaries. The useful question is not “Which protocol wins?” but “Which relationship in the system needs to be standardized?”\"},\"tunes\":{}},{\"id\":\"direct\",\"type\":\"callout\",\"data\":{\"variant\":\"info\",\"title\":\"Direct answer\",\"body\":\"\u003Cstrong>MCP, A2A, UCP, AP2 and A2UI are mostly complementary, not substitutes.\u003C\u002Fstrong> MCP connects an AI application to tools, data and resources. A2A connects independent agents to one another. UCP standardizes commerce interactions between consumer surfaces, businesses and payment ecosystems. AP2 adds verifiable authorization and payment intent to agent-led transactions. A2UI lets an agent describe interactive UI without sending arbitrary application code. A production system may legitimately use several of them in one workflow.\"},\"tunes\":{}},{\"id\":\"freshness\",\"type\":\"callout\",\"data\":{\"variant\":\"warning\",\"title\":\"Fast-moving standards\",\"body\":\"This article reflects the protocol landscape on \u003Cstrong>25 September 2026\u003C\u002Fstrong>. A2A has reached v1.0, MCP's current TypeScript v2 line implements the 2026-07-28 specification, UCP has already added newer 2026 protocol versions, and A2UI continues to evolve. Always verify the current specification before implementation.\"},\"tunes\":{}},{\"id\":\"model-note\",\"type\":\"callout\",\"data\":{\"variant\":\"note\",\"title\":\"The model used in this article\",\"body\":\"The Protocol Responsibility Stack and Protocol Selection Test below are practical architecture models proposed here. They are not official terminology from the protocol projects.\"},\"tunes\":{}},{\"id\":\"h-core\",\"type\":\"header\",\"data\":{\"text\":\"The core mistake: comparing protocols that sit at different boundaries\",\"level\":2},\"tunes\":{}},{\"id\":\"p-core-1\",\"type\":\"paragraph\",\"data\":{\"text\":\"A protocol is useful because two independently implemented systems need a stable contract. The contract only makes sense if the boundary is clear. An agent talking to a database has a different interoperability problem from one agent delegating work to another, a shopper authorizing a purchase, or a remote agent asking a native application to render a form.\"},\"tunes\":{}},{\"id\":\"p-core-2\",\"type\":\"paragraph\",\"data\":{\"text\":\"Google's 2026 developer guide explicitly presents MCP, A2A, UCP, AP2, A2UI and related UI protocols as a stack of complementary standards. The same example workflow can use several of them together: tools for inventory, remote agents for suppliers, commerce for ordering, payment authorization for spending and UI protocols for interaction.\"},\"tunes\":{}},{\"id\":\"h-stack\",\"type\":\"header\",\"data\":{\"text\":\"The Protocol Responsibility Stack\",\"level\":2},\"tunes\":{}},{\"id\":\"stack-table\",\"type\":\"table\",\"data\":{\"withHeadings\":true,\"stretched\":false,\"content\":[[\"Protocol\",\"Standardizes which relationship?\",\"Primary abstraction\",\"Not primarily for\"],[\"MCP\",\"AI application ↔ tools, resources and data\",\"Tools, resources, prompts and host\u002Fserver capability exchange\",\"Independent agent collaboration or commerce semantics\"],[\"A2A\",\"Agent ↔ independent agent\",\"Agent discovery, messages, tasks, artifacts and long-running collaboration\",\"Direct database\u002Ftool integration\"],[\"UCP\",\"Consumer\u002Fagent surface ↔ merchant commerce system\",\"Product\u002Fcart\u002Fcheckout\u002Ffulfillment\u002Forder capabilities\",\"General-purpose agent communication\"],[\"AP2\",\"User\u002Fagent intent ↔ payment authorization\",\"Mandates, approval constraints and auditable agent-led payment authority\",\"Product discovery or generic checkout transport\"],[\"A2UI\",\"Agent ↔ user interface host\",\"Declarative UI intent rendered by trusted native components\",\"Arbitrary remote frontend code or agent-to-agent task delegation\"]]},\"tunes\":{}},{\"id\":\"all-five\",\"type\":\"callout\",\"data\":{\"variant\":\"success\",\"title\":\"One workflow can use all five\",\"body\":\"An agent may use \u003Cstrong>MCP\u003C\u002Fstrong> to inspect inventory, \u003Cstrong>A2A\u003C\u002Fstrong> to ask a supplier agent for availability, \u003Cstrong>UCP\u003C\u002Fstrong> to build a commerce transaction, \u003Cstrong>AP2\u003C\u002Fstrong> to prove spending authority, and \u003Cstrong>A2UI\u003C\u002Fstrong> to render a native approval interface to the user.\"},\"tunes\":{}},{\"id\":\"h-mcp\",\"type\":\"header\",\"data\":{\"text\":\"1. MCP: connect the agent to capabilities\",\"level\":2},\"tunes\":{}},{\"id\":\"p-mcp-1\",\"type\":\"paragraph\",\"data\":{\"text\":\"The Model Context Protocol is an open standard for connecting AI applications to external systems where tools, data and reusable resources live. A server exposes capabilities; an MCP host connects to that server and makes those capabilities available to the model or application.\"},\"tunes\":{}},{\"id\":\"p-mcp-2\",\"type\":\"paragraph\",\"data\":{\"text\":\"The current MCP TypeScript v2 documentation describes the protocol in exactly those terms: servers expose tools, resources and prompts, while hosts such as development environments or custom applications connect to them. This makes MCP primarily a capability integration protocol.\"},\"tunes\":{}},{\"id\":\"h-mcp-use\",\"type\":\"header\",\"data\":{\"text\":\"Use MCP when\",\"level\":3},\"tunes\":{}},{\"id\":\"mcp-list\",\"type\":\"list\",\"data\":{\"style\":\"unordered\",\"meta\":{},\"items\":[\"An AI application needs standardized access to tools or APIs.\",\"You want one capability server to work with multiple compatible AI hosts.\",\"You need structured access to data or resources without hard-coding every integration into each agent.\",\"The external system is a capability provider, not an autonomous peer agent.\"]},\"tunes\":{}},{\"id\":\"mcp-warning\",\"type\":\"callout\",\"data\":{\"variant\":\"warning\",\"title\":\"MCP is not automatically A2A\",\"body\":\"An MCP server can expose powerful functions, but that does not make it an independent agent with its own task lifecycle, discovery semantics and opaque internal reasoning. Tool invocation and agent collaboration are different contracts.\"},\"tunes\":{}},{\"id\":\"h-a2a\",\"type\":\"header\",\"data\":{\"text\":\"2. A2A: connect independent agents\",\"level\":2},\"tunes\":{}},{\"id\":\"p-a2a-1\",\"type\":\"paragraph\",\"data\":{\"text\":\"Agent2Agent (A2A) is designed for communication between independent, potentially opaque agent systems. Its current v1.0 specification focuses on capability discovery, messaging, tasks, artifacts, multimodal content and long-running collaboration without requiring one agent to expose its internal tools, memory or implementation to another.\"},\"tunes\":{}},{\"id\":\"p-a2a-2\",\"type\":\"paragraph\",\"data\":{\"text\":\"That opacity is the important boundary. The calling agent does not need to know whether the remote agent uses MCP, custom tools, a proprietary planner, another model vendor, or human escalation internally. It needs a contract for discovering capabilities and delegating work.\"},\"tunes\":{}},{\"id\":\"p-a2a-3\",\"type\":\"paragraph\",\"data\":{\"text\":\"A2A v1.0 also standardizes version negotiation and supports multiple bindings around a common data model. Its published Agent Card mechanism gives clients a standard discovery point for an agent's capabilities, supported protocols, authentication requirements and skills.\"},\"tunes\":{}},{\"id\":\"h-a2a-use\",\"type\":\"header\",\"data\":{\"text\":\"Use A2A when\",\"level\":3},\"tunes\":{}},{\"id\":\"a2a-list\",\"type\":\"list\",\"data\":{\"style\":\"unordered\",\"meta\":{},\"items\":[\"One autonomous agent needs to delegate work to another autonomous agent.\",\"The remote system should remain opaque behind a capability contract.\",\"Tasks may be long-running, asynchronous or require human-in-the-loop interaction.\",\"Agents are built with different frameworks, languages, vendors or organizational ownership.\"]},\"tunes\":{}},{\"id\":\"h-mcp-a2a\",\"type\":\"header\",\"data\":{\"text\":\"MCP vs A2A: vertical integration vs horizontal collaboration\",\"level\":2},\"tunes\":{}},{\"id\":\"mcp-a2a-comparison\",\"type\":\"comparison\",\"data\":{\"title\":\"MCP and A2A solve different interoperability problems\",\"layout\":\"table\",\"columns\":[{\"id\":\"dimension\",\"label\":\"Dimension\"},{\"id\":\"mcp\",\"label\":\"MCP\"},{\"id\":\"a2a\",\"label\":\"A2A\"}],\"rows\":[{\"id\":\"relationship\",\"label\":\"Relationship\",\"values\":[\"\",\"\",\"\"]},{\"id\":\"abstraction\",\"label\":\"Abstraction\",\"values\":[\"\",\"\",\"\"]},{\"id\":\"opacity\",\"label\":\"Internal opacity\",\"values\":[\"\",\"\",\"\"]},{\"id\":\"long\",\"label\":\"Long-running work\",\"values\":[\"\",\"\",\"\"]}]},\"tunes\":{}},{\"id\":\"p-mcp-a2a-1\",\"type\":\"paragraph\",\"data\":{\"text\":\"The A2A project itself now describes the distinction as horizontal versus vertical: MCP connects agents to internal tools and databases, while A2A enables peer-to-peer collaboration across agent systems.\"},\"tunes\":{}},{\"id\":\"h-ucp\",\"type\":\"header\",\"data\":{\"text\":\"3. UCP: standardize agentic commerce\",\"level\":2},\"tunes\":{}},{\"id\":\"p-ucp-1\",\"type\":\"paragraph\",\"data\":{\"text\":\"The Universal Commerce Protocol is not a generic agent protocol. It standardizes commerce journeys between consumer surfaces, merchants and payment providers. Google's implementation already supports capabilities such as cart creation, checkout, fulfillment and order lifecycle through versioned profiles and APIs.\"},\"tunes\":{}},{\"id\":\"p-ucp-2\",\"type\":\"paragraph\",\"data\":{\"text\":\"A merchant can publish a UCP profile under \u002F.well-known\u002Fucp describing services, protocol versions and capabilities. That discovery pattern matters because an agentic surface should not need a bespoke checkout contract for every merchant.\"},\"tunes\":{}},{\"id\":\"p-ucp-3\",\"type\":\"paragraph\",\"data\":{\"text\":\"UCP is also intentionally composable. Google's technical overview says it can integrate through APIs, A2A and MCP and is compatible with AP2 for agentic payment authorization.\"},\"tunes\":{}},{\"id\":\"h-ucp-use\",\"type\":\"header\",\"data\":{\"text\":\"Use UCP when\",\"level\":3},\"tunes\":{}},{\"id\":\"ucp-list\",\"type\":\"list\",\"data\":{\"style\":\"unordered\",\"meta\":{},\"items\":[\"The workflow involves merchant products, carts, checkout, fulfillment or order lifecycle.\",\"You are building a merchant surface that should work with agentic shopping experiences.\",\"The integration needs commerce-specific semantics rather than generic tool calls.\",\"You want an interoperable commerce contract that can coexist with MCP, A2A and payment protocols.\"]},\"tunes\":{}},{\"id\":\"h-ap2\",\"type\":\"header\",\"data\":{\"text\":\"4. AP2: prove that the agent was allowed to spend\",\"level\":2},\"tunes\":{}},{\"id\":\"p-ap2-1\",\"type\":\"paragraph\",\"data\":{\"text\":\"Agentic commerce introduces a problem that ordinary checkout flows did not have to solve in the same way: an agent may transact when the human is not clicking the final button in real time. The Agent Payments Protocol (AP2) addresses authorization, authenticity and accountability for agent-led payments.\"},\"tunes\":{}},{\"id\":\"p-ap2-2\",\"type\":\"paragraph\",\"data\":{\"text\":\"Google's 2026 protocol guide describes AP2 through typed mandates that capture user intent, spending constraints and the specific transaction being authorized. AP2 can work as an extension alongside UCP: UCP describes the commerce transaction, while AP2 provides evidence that the agent had authority to perform the payment.\"},\"tunes\":{}},{\"id\":\"p-ap2-3\",\"type\":\"paragraph\",\"data\":{\"text\":\"This distinction is important. A checkout protocol can tell a merchant what should be purchased. It does not by itself prove who authorized the agent to spend, under what limit, for which merchant, for how long, or whether the final cart remained inside that authority.\"},\"tunes\":{}},{\"id\":\"h-ucp-ap2\",\"type\":\"header\",\"data\":{\"text\":\"UCP vs AP2: transaction semantics vs authority\",\"level\":2},\"tunes\":{}},{\"id\":\"ucp-ap2-table\",\"type\":\"table\",\"data\":{\"withHeadings\":true,\"stretched\":false,\"content\":[[\"Question\",\"UCP\",\"AP2\"],[\"What is being bought?\",\"Commerce items, cart, checkout and fulfillment semantics\",\"References the authorized transaction context\"],[\"Who may authorize it?\",\"Not the primary protocol responsibility\",\"Explicit agent\u002Fuser authority and mandate model\"],[\"What spending constraints apply?\",\"Commerce flow can contain totals and checkout data\",\"Authorization guardrails and intent limits\"],[\"How is the transaction audited?\",\"Order and commerce lifecycle\",\"Cryptographic \u002F verifiable authorization trail through mandates and receipts\"],[\"Can they work together?\",\"Yes\",\"Yes — AP2 can extend agentic commerce flows\"]]},\"tunes\":{}},{\"id\":\"h-a2ui\",\"type\":\"header\",\"data\":{\"text\":\"5. A2UI: let agents describe interfaces without owning your frontend\",\"level\":2},\"tunes\":{}},{\"id\":\"p-a2ui-1\",\"type\":\"paragraph\",\"data\":{\"text\":\"Agent-to-User Interface (A2UI) tackles another boundary: how a remote or local agent communicates a rich interactive interface to a host application. Instead of sending arbitrary HTML, CSS and JavaScript, A2UI uses declarative data that the host renders through its own trusted component catalog.\"},\"tunes\":{}},{\"id\":\"p-a2ui-2\",\"type\":\"paragraph\",\"data\":{\"text\":\"This preserves the host application's design system and security model while still allowing an agent to request dynamic interfaces. A2UI v0.9 specifically emphasizes framework-agnostic UI intent and streaming updates across web, mobile and other clients.\"},\"tunes\":{}},{\"id\":\"p-a2ui-3\",\"type\":\"paragraph\",\"data\":{\"text\":\"Google's later A2UI + MCP Apps work also demonstrates that these UI models are not necessarily mutually exclusive. Declarative native UI and richer embedded application experiences can coexist depending on the task.\"},\"tunes\":{}},{\"id\":\"h-a2ui-use\",\"type\":\"header\",\"data\":{\"text\":\"Use A2UI when\",\"level\":3},\"tunes\":{}},{\"id\":\"a2ui-list\",\"type\":\"list\",\"data\":{\"style\":\"unordered\",\"meta\":{},\"items\":[\"A remote agent needs to request forms, cards, controls or other interactive UI.\",\"The host should preserve its native components, styling and security boundary.\",\"You do not want remote agents shipping arbitrary executable frontend code.\",\"The same agent-defined UI intent should work across different client frameworks.\"]},\"tunes\":{}},{\"id\":\"h-selection\",\"type\":\"header\",\"data\":{\"text\":\"The Protocol Selection Test\",\"level\":2},\"tunes\":{}},{\"id\":\"p-selection-intro\",\"type\":\"paragraph\",\"data\":{\"text\":\"Do not start from the acronym. Start from the relationship that needs interoperability.\"},\"tunes\":{}},{\"id\":\"selection-flow\",\"type\":\"processFlow\",\"data\":{\"title\":\"Choose the protocol by the boundary\",\"orientation\":\"auto\",\"steps\":[{\"label\":\"1. Identify the two independent parties\",\"description\":\"Is this AI-to-tool, agent-to-agent, agent-to-merchant, agent-to-payment authority, or agent-to-user-interface?\"},{\"label\":\"2. Identify the shared object\",\"description\":\"Is the contract about a tool call, task, cart, payment mandate, artifact, or UI description?\"},{\"label\":\"3. Check whether a domain protocol already exists\",\"description\":\"Prefer commerce or payment semantics when the problem is commerce or authorization instead of encoding everything as generic tools.\"},{\"label\":\"4. Keep local internals local\",\"description\":\"Do not expose an entire agent as MCP tools if the remote party only needs an A2A capability, and do not make a remote agent responsible for your UI runtime.\"},{\"label\":\"5. Compose protocols when the workflow crosses boundaries\",\"description\":\"One workflow can legitimately cross tool, agent, commerce, payment and UI contracts.\"},{\"label\":\"6. Version each contract independently\",\"description\":\"Protocol versions evolve at different speeds; do not tie every integration to one monolithic application version.\"},{\"label\":\"7. Preserve authorization at every boundary\",\"description\":\"Interoperability does not replace product permissions, tool authorization, payment authority or data-access policy.\"}]},\"tunes\":{}},{\"id\":\"h-workflow\",\"type\":\"header\",\"data\":{\"text\":\"A realistic multi-protocol workflow\",\"level\":2},\"tunes\":{}},{\"id\":\"workflow-flow\",\"type\":\"processFlow\",\"data\":{\"title\":\"Example: an autonomous procurement workflow\",\"orientation\":\"auto\",\"steps\":[{\"label\":\"1. Inspect internal stock with MCP\",\"description\":\"The purchasing agent calls inventory and forecasting capabilities exposed by internal MCP servers.\"},{\"label\":\"2. Discover a supplier agent with A2A\",\"description\":\"The agent reads the supplier's Agent Card and delegates an availability and lead-time task.\"},{\"label\":\"3. Negotiate the commerce object with UCP\",\"description\":\"The supplier or merchant surface returns structured cart, checkout and fulfillment information.\"},{\"label\":\"4. Check spending authority with AP2\",\"description\":\"The purchase is compared with the user's or organization's signed mandate, merchant constraints and spending limits.\"},{\"label\":\"5. Ask for approval through A2UI\",\"description\":\"If human approval is required, the agent sends declarative UI intent and the host renders the approval experience using trusted native components.\"},{\"label\":\"6. Complete and audit\",\"description\":\"Commerce state, payment authorization, agent task evidence and application audit records remain traceable across their respective boundaries.\"}]},\"tunes\":{}},{\"id\":\"h-universal\",\"type\":\"header\",\"data\":{\"text\":\"Why one universal agent protocol is unlikely to replace all of them\",\"level\":2},\"tunes\":{}},{\"id\":\"p-universal-1\",\"type\":\"paragraph\",\"data\":{\"text\":\"A universal protocol sounds simpler until it must encode every domain's semantics. Tool discovery, long-running agent collaboration, checkout, payment authorization and native UI all have different lifecycle, security and correctness requirements.\"},\"tunes\":{}},{\"id\":\"p-universal-2\",\"type\":\"paragraph\",\"data\":{\"text\":\"The web itself evolved through layered protocols rather than one message format for every problem. The emerging agentic stack appears to be moving in the same direction: common horizontal primitives, specialized domain contracts and explicit discovery\u002Fversioning.\"},\"tunes\":{}},{\"id\":\"p-universal-3\",\"type\":\"paragraph\",\"data\":{\"text\":\"The architecture challenge therefore shifts from “which protocol wins?” to how cleanly protocols compose without duplicating identity, authorization, state and audit semantics.\"},\"tunes\":{}},{\"id\":\"h-failures\",\"type\":\"header\",\"data\":{\"text\":\"Protocol composition creates new failure modes\",\"level\":2},\"tunes\":{}},{\"id\":\"failure-table\",\"type\":\"table\",\"data\":{\"withHeadings\":true,\"stretched\":false,\"content\":[[\"Failure mode\",\"What happens\",\"Architecture control\"],[\"Authority leakage\",\"A valid tool or agent capability is treated as permission to perform a business action\",\"Keep product authorization independent from protocol capability discovery\"],[\"Identity mismatch\",\"MCP host identity, A2A agent identity and commerce\u002Fpayment identity refer to different principals\",\"Define explicit principal mapping across boundaries\"],[\"Version drift\",\"One protocol upgrades while dependent adapters assume older semantics\",\"Negotiate and pin protocol versions independently\"],[\"State duplication\",\"The same cart, task or approval state is copied into several protocol layers\",\"Define one authoritative owner per domain object\"],[\"Audit fragmentation\",\"Tool traces, agent tasks, checkout and payment evidence cannot be joined\",\"Carry correlation IDs and stable domain identifiers across protocol boundaries\"],[\"Semantic tunneling\",\"Everything is forced through a generic protocol as opaque JSON\",\"Use domain protocols where their semantics materially improve correctness\"]]},\"tunes\":{}},{\"id\":\"h-app-arch\",\"type\":\"header\",\"data\":{\"text\":\"Protocol choice does not replace application architecture\",\"level\":2},\"tunes\":{}},{\"id\":\"p-app-1\",\"type\":\"paragraph\",\"data\":{\"text\":\"Open standards reduce integration coupling, but they do not decide your domain model, authorization policy, source of truth, retry strategy or acceptance criteria. An MCP tool can still expose the wrong capability. An A2A agent can still return a bad artifact. A UCP checkout can still contain stale merchant data. An AP2 mandate can still be misapplied by application logic.\"},\"tunes\":{}},{\"id\":\"p-app-2\",\"type\":\"paragraph\",\"data\":{\"text\":\"Treat protocols as contracts between independently evolving components. Keep domain truth and consequential policy in the application layer that owns them, then use protocols to make the boundaries interoperable.\"},\"tunes\":{}},{\"id\":\"ref-harness\",\"type\":\"referralArticle\",\"data\":{\"url\":\"https:\u002F\u002Fstajic.de\u002Fblog\u002Fmanaged-agent-harness-vs-self-hosted-agent-loop-what-you-gain-what-you-lose\",\"title\":\"Managed Agent Harness vs Self-Hosted Agent Loop: What You Gain, What You Lose\",\"excerpt\":\"Protocol boundaries solve interoperability. Runtime boundaries solve who operates the harness, execution environment and application control plane.\",\"ctaLabel\":\"Read the runtime architecture guide\"},\"tunes\":{}},{\"id\":\"h-change\",\"type\":\"header\",\"data\":{\"text\":\"What would change this answer?\",\"level\":2},\"tunes\":{}},{\"id\":\"p-change-1\",\"type\":\"paragraph\",\"data\":{\"text\":\"The stack changes if protocols converge, one standard formally absorbs another, or vendors standardize a shared identity and authorization layer across several boundaries. UCP already demonstrates composition by supporting APIs, A2A and MCP and by integrating with AP2 rather than replacing them.\"},\"tunes\":{}},{\"id\":\"p-change-2\",\"type\":\"paragraph\",\"data\":{\"text\":\"The answer also changes by application scope. A small internal agent may need only MCP. A multi-company workflow may need A2A. A merchant may need UCP without A2UI. A delegated purchasing agent may need all of them. Use the smallest protocol set that represents the real boundaries without flattening domain semantics.\"},\"tunes\":{}},{\"id\":\"h-limitations\",\"type\":\"header\",\"data\":{\"text\":\"Limitations\",\"level\":2},\"tunes\":{}},{\"id\":\"p-limit-1\",\"type\":\"paragraph\",\"data\":{\"text\":\"The protocols discussed here are at different maturity levels and have different governance models. A2A has reached a stable v1.0 specification, while other standards continue to evolve rapidly. Ecosystem adoption is also uneven across vendors and frameworks.\"},\"tunes\":{}},{\"id\":\"p-limit-2\",\"type\":\"paragraph\",\"data\":{\"text\":\"This article focuses on architecture responsibility rather than implementation completeness. Specific authentication methods, transport bindings, schemas and extension mechanisms must be taken from each protocol's current specification.\"},\"tunes\":{}},{\"id\":\"h-conclusion\",\"type\":\"header\",\"data\":{\"text\":\"Conclusion\",\"level\":2},\"tunes\":{}},{\"id\":\"p-conclusion-1\",\"type\":\"paragraph\",\"data\":{\"text\":\"MCP, A2A, UCP, AP2 and A2UI make more sense when viewed as protocols for different relationships, not five competing attempts to standardize “agents.”\"},\"tunes\":{}},{\"id\":\"p-conclusion-2\",\"type\":\"paragraph\",\"data\":{\"text\":\"MCP exposes capabilities. A2A coordinates independent agents. UCP gives commerce its own machine-readable contract. AP2 adds verifiable payment authority. A2UI gives agents a safe declarative path into user interfaces. The emerging agentic web is therefore not replacing protocols with AI; it is creating a new protocol stack around AI.\"},\"tunes\":{}},{\"id\":\"h-faq\",\"type\":\"header\",\"data\":{\"text\":\"FAQ\",\"level\":2},\"tunes\":{}},{\"id\":\"faq\",\"type\":\"faq\",\"data\":{\"title\":\"MCP, A2A, UCP, AP2 and A2UI\",\"items\":[{\"id\":\"faq1\",\"question\":\"Is A2A a replacement for MCP?\",\"answer\":\"No. MCP primarily standardizes how AI applications access tools, resources and data. A2A standardizes collaboration between independent agent systems. A remote agent can internally use MCP while exposing an A2A interface.\"},{\"id\":\"faq2\",\"question\":\"Is UCP a replacement for MCP in shopping agents?\",\"answer\":\"Not generally. UCP provides commerce-specific semantics such as cart, checkout and fulfillment. MCP can still expose merchant tools or data, and UCP is designed to coexist with MCP and A2A.\"},{\"id\":\"faq3\",\"question\":\"What is the difference between UCP and AP2?\",\"answer\":\"UCP standardizes commerce interactions and transaction lifecycle. AP2 focuses on proving that an agent had authority to perform a payment under defined user or organizational constraints.\"},{\"id\":\"faq4\",\"question\":\"What problem does A2UI solve?\",\"answer\":\"A2UI lets agents send declarative UI intent to a host application, which renders the experience through trusted native components instead of executing arbitrary remote frontend code.\"},{\"id\":\"faq5\",\"question\":\"Can one agent application use all of these protocols?\",\"answer\":\"Yes. A workflow can use MCP for internal tools, A2A for remote-agent delegation, UCP for commerce, AP2 for payment authorization and A2UI for human interaction.\"},{\"id\":\"faq6\",\"question\":\"Which protocol should I implement first?\",\"answer\":\"Start from the interoperability boundary. If the problem is tool access, evaluate MCP. If it is independent-agent collaboration, evaluate A2A. If it is commerce, UCP. If it is delegated payment authority, AP2. If it is portable agent-driven UI, A2UI.\"}]},\"tunes\":{}},{\"id\":\"h-glossary\",\"type\":\"header\",\"data\":{\"text\":\"Glossary\",\"level\":2},\"tunes\":{}},{\"id\":\"glossary\",\"type\":\"glossary\",\"data\":{\"title\":\"Key agent-protocol terms\",\"entries\":[{\"term\":\"MCP\",\"definition\":\"Model Context Protocol, an open standard for exposing tools, resources and prompts from external systems to compatible AI hosts.\",\"anchor\":\"mcp\"},{\"term\":\"A2A\",\"definition\":\"Agent2Agent Protocol, an open standard for discovering and collaborating with independent agent systems through messages, tasks and artifacts.\",\"anchor\":\"a2a\"},{\"term\":\"UCP\",\"definition\":\"Universal Commerce Protocol, an open standard for interoperable agentic commerce journeys between consumer surfaces, businesses and payment providers.\",\"anchor\":\"ucp\"},{\"term\":\"AP2\",\"definition\":\"Agent Payments Protocol, an open standard for representing and verifying authority, intent and accountability in agent-led payments.\",\"anchor\":\"ap2\"},{\"term\":\"A2UI\",\"definition\":\"Agent-to-User Interface, a declarative protocol for allowing agents to request UI that is rendered using the host application's trusted components.\",\"anchor\":\"a2ui\"},{\"term\":\"Protocol composition\",\"definition\":\"Using multiple protocols in one workflow, each responsible for a distinct interoperability boundary rather than forcing all semantics through one contract.\",\"anchor\":\"protocol-composition\"}]},\"tunes\":{}},{\"id\":\"h-sources\",\"type\":\"header\",\"data\":{\"text\":\"Primary sources and further reading\",\"level\":2},\"tunes\":{}},{\"id\":\"src-google-guide\",\"type\":\"linkTool\",\"data\":{\"link\":\"https:\u002F\u002Fdevelopers.googleblog.com\u002Fdevelopers-guide-to-ai-agent-protocols\u002F\",\"meta\":{\"image\":{\"url\":\"\"},\"title\":\"Google Developers — Developer's Guide to AI Agent Protocols\",\"description\":\"A practical overview showing MCP, A2A, UCP, AP2, A2UI and related protocols working together in one multi-step agent workflow.\"}},\"tunes\":{}},{\"id\":\"src-mcp\",\"type\":\"linkTool\",\"data\":{\"link\":\"https:\u002F\u002Fts.sdk.modelcontextprotocol.io\u002Fv2\u002F\",\"meta\":{\"image\":{\"url\":\"\"},\"title\":\"Model Context Protocol — TypeScript SDK v2\",\"description\":\"Current stable SDK documentation implementing the 2026-07-28 MCP specification and defining tools, resources, prompts and host\u002Fserver integration.\"}},\"tunes\":{}},{\"id\":\"src-a2a-spec\",\"type\":\"linkTool\",\"data\":{\"link\":\"https:\u002F\u002Fa2a-protocol.org\u002Fdev\u002Fspecification\u002F\",\"meta\":{\"image\":{\"url\":\"\"},\"title\":\"A2A Protocol — v1.0 Specification\",\"description\":\"Current A2A protocol specification covering Agent Cards, messages, tasks, artifacts, bindings and version negotiation.\"}},\"tunes\":{}},{\"id\":\"src-a2a-aaif\",\"type\":\"linkTool\",\"data\":{\"link\":\"https:\u002F\u002Fa2a-protocol.org\u002Flatest\u002Fblog\u002F2026\u002F08\u002F27\u002Fa-new-chapter-for-a2a-joining-the-agentic-ai-foundation\u002F\",\"meta\":{\"image\":{\"url\":\"\"},\"title\":\"A2A — Joining the Agentic AI Foundation\",\"description\":\"Current project framing of A2A as the horizontal agent-collaboration layer alongside MCP as vertical tool\u002Fdata integration.\"}},\"tunes\":{}},{\"id\":\"src-ucp\",\"type\":\"linkTool\",\"data\":{\"link\":\"https:\u002F\u002Fdevelopers.googleblog.com\u002Funder-the-hood-universal-commerce-protocol-ucp\u002F\",\"meta\":{\"image\":{\"url\":\"\"},\"title\":\"Google Developers — Under the Hood: Universal Commerce Protocol\",\"description\":\"Technical overview of UCP, its commerce primitives and its ability to compose with APIs, A2A, MCP and AP2.\"}},\"tunes\":{}},{\"id\":\"src-ucp-profile\",\"type\":\"linkTool\",\"data\":{\"link\":\"https:\u002F\u002Fdevelopers.google.com\u002Fmerchant\u002Fucp\u002Fimplementation\u002F2026-04-08\u002Fpublish-profile\",\"meta\":{\"image\":{\"url\":\"\"},\"title\":\"Google Universal Commerce Protocol — UCP Profile\",\"description\":\"Current versioned profile mechanism for publishing UCP services and merchant commerce capabilities.\"}},\"tunes\":{}},{\"id\":\"src-ap2\",\"type\":\"linkTool\",\"data\":{\"link\":\"https:\u002F\u002Fcloud.google.com\u002Fblog\u002Fproducts\u002Fai-machine-learning\u002Fannouncing-agents-to-payments-ap2-protocol\",\"meta\":{\"image\":{\"url\":\"\"},\"title\":\"Google Cloud — Agent Payments Protocol (AP2)\",\"description\":\"Announcement and rationale for an open protocol covering authorization, authenticity and accountability in agent-led payments.\"}},\"tunes\":{}},{\"id\":\"src-a2ui\",\"type\":\"linkTool\",\"data\":{\"link\":\"https:\u002F\u002Fdevelopers.googleblog.com\u002Fa2ui-v0-9-generative-ui\u002F\",\"meta\":{\"image\":{\"url\":\"\"},\"title\":\"Google Developers — A2UI v0.9\",\"description\":\"A2UI's framework-agnostic declarative model for portable agent-driven interfaces rendered by host-native components.\"}},\"tunes\":{}},{\"id\":\"src-a2ui-mcp\",\"type\":\"linkTool\",\"data\":{\"link\":\"https:\u002F\u002Fdevelopers.googleblog.com\u002Fa2ui-and-mcp-apps\u002F\",\"meta\":{\"image\":{\"url\":\"\"},\"title\":\"Google Developers — A2UI + MCP Apps\",\"description\":\"How declarative A2UI and richer MCP App experiences can coexist rather than being treated as mutually exclusive UI models.\"}},\"tunes\":{}}],\"version\":\"2.31.6\"}",{"time":936,"blocks":937,"version":1484},1790352626854,[938,942,946,951,956,961,965,969,973,977,1006,1011,1015,1019,1023,1027,1036,1041,1045,1049,1053,1057,1061,1070,1074,1096,1100,1104,1108,1112,1116,1120,1129,1133,1137,1141,1145,1149,1175,1179,1183,1187,1191,1195,1204,1208,1212,1238,1242,1265,1269,1273,1277,1281,1285,1317,1321,1325,1329,1336,1340,1344,1348,1352,1356,1360,1364,1368,1372,1376,1399,1403,1421,1425,1432,1438,1445,1452,1459,1466,1472,1478],{"id":215,"data":939,"type":220,"tunes":941},{"title":940,"maxLevel":218,"minLevel":219},"Contents",{},{"id":223,"data":943,"type":226,"tunes":945},{"text":944},"AI agent protocols are multiplying quickly: MCP, A2A, UCP, AP2, A2UI and adjacent standards increasingly appear in the same architecture diagrams. They are often described as competing protocols. In practice, most of them solve different interoperability problems at different boundaries. The useful question is not “Which protocol wins?” but “Which relationship in the system needs to be standardized?”",{},{"id":229,"data":947,"type":234,"tunes":950},{"body":948,"title":949,"variant":233},"\u003Cstrong>MCP, A2A, UCP, AP2 and A2UI are mostly complementary, not substitutes.\u003C\u002Fstrong> MCP connects an AI application to tools, data and resources. A2A connects independent agents to one another. UCP standardizes commerce interactions between consumer surfaces, businesses and payment ecosystems. AP2 adds verifiable authorization and payment intent to agent-led transactions. A2UI lets an agent describe interactive UI without sending arbitrary application code. A production system may legitimately use several of them in one workflow.","Direct answer",{},{"id":237,"data":952,"type":234,"tunes":955},{"body":953,"title":954,"variant":241},"This article reflects the protocol landscape on \u003Cstrong>25 September 2026\u003C\u002Fstrong>. A2A has reached v1.0, MCP's current TypeScript v2 line implements the 2026-07-28 specification, UCP has already added newer 2026 protocol versions, and A2UI continues to evolve. Always verify the current specification before implementation.","Fast-moving standards",{},{"id":244,"data":957,"type":234,"tunes":960},{"body":958,"title":959,"variant":248},"The Protocol Responsibility Stack and Protocol Selection Test below are practical architecture models proposed here. They are not official terminology from the protocol projects.","The model used in this article",{},{"id":251,"data":962,"type":42,"tunes":964},{"text":963,"level":219},"The core mistake: comparing protocols that sit at different boundaries",{},{"id":256,"data":966,"type":226,"tunes":968},{"text":967},"A protocol is useful because two independently implemented systems need a stable contract. The contract only makes sense if the boundary is clear. An agent talking to a database has a different interoperability problem from one agent delegating work to another, a shopper authorizing a purchase, or a remote agent asking a native application to render a form.",{},{"id":261,"data":970,"type":226,"tunes":972},{"text":971},"Google's 2026 developer guide explicitly presents MCP, A2A, UCP, AP2, A2UI and related UI protocols as a stack of complementary standards. The same example workflow can use several of them together: tools for inventory, remote agents for suppliers, commerce for ordering, payment authorization for spending and UI protocols for interaction.",{},{"id":266,"data":974,"type":42,"tunes":976},{"text":975,"level":219},"The Protocol Responsibility Stack",{},{"id":271,"data":978,"type":304,"tunes":1005},{"content":979,"stretched":43,"withHeadings":14},[980,985,989,993,997,1001],[981,982,983,984],"Protocol","Standardizes which relationship?","Primary abstraction","Not primarily for",[280,986,987,988],"AI application ↔ tools, resources and data","Tools, resources, prompts and host\u002Fserver capability exchange","Independent agent collaboration or commerce semantics",[285,990,991,992],"Agent ↔ independent agent","Agent discovery, messages, tasks, artifacts and long-running collaboration","Direct database\u002Ftool integration",[290,994,995,996],"Consumer\u002Fagent surface ↔ merchant commerce system","Product\u002Fcart\u002Fcheckout\u002Ffulfillment\u002Forder capabilities","General-purpose agent communication",[295,998,999,1000],"User\u002Fagent intent ↔ payment authorization","Mandates, approval constraints and auditable agent-led payment authority","Product discovery or generic checkout transport",[300,1002,1003,1004],"Agent ↔ user interface host","Declarative UI intent rendered by trusted native components","Arbitrary remote frontend code or agent-to-agent task delegation",{},{"id":307,"data":1007,"type":234,"tunes":1010},{"body":1008,"title":1009,"variant":311},"An agent may use \u003Cstrong>MCP\u003C\u002Fstrong> to inspect inventory, \u003Cstrong>A2A\u003C\u002Fstrong> to ask a supplier agent for availability, \u003Cstrong>UCP\u003C\u002Fstrong> to build a commerce transaction, \u003Cstrong>AP2\u003C\u002Fstrong> to prove spending authority, and \u003Cstrong>A2UI\u003C\u002Fstrong> to render a native approval interface to the user.","One workflow can use all five",{},{"id":314,"data":1012,"type":42,"tunes":1014},{"text":1013,"level":219},"1. MCP: connect the agent to capabilities",{},{"id":319,"data":1016,"type":226,"tunes":1018},{"text":1017},"The Model Context Protocol is an open standard for connecting AI applications to external systems where tools, data and reusable resources live. A server exposes capabilities; an MCP host connects to that server and makes those capabilities available to the model or application.",{},{"id":324,"data":1020,"type":226,"tunes":1022},{"text":1021},"The current MCP TypeScript v2 documentation describes the protocol in exactly those terms: servers expose tools, resources and prompts, while hosts such as development environments or custom applications connect to them. This makes MCP primarily a capability integration protocol.",{},{"id":329,"data":1024,"type":42,"tunes":1026},{"text":1025,"level":218},"Use MCP when",{},{"id":334,"data":1028,"type":343,"tunes":1035},{"meta":1029,"items":1030,"style":342},{},[1031,1032,1033,1034],"An AI application needs standardized access to tools or APIs.","You want one capability server to work with multiple compatible AI hosts.","You need structured access to data or resources without hard-coding every integration into each agent.","The external system is a capability provider, not an autonomous peer agent.",{},{"id":346,"data":1037,"type":234,"tunes":1040},{"body":1038,"title":1039,"variant":241},"An MCP server can expose powerful functions, but that does not make it an independent agent with its own task lifecycle, discovery semantics and opaque internal reasoning. Tool invocation and agent collaboration are different contracts.","MCP is not automatically A2A",{},{"id":352,"data":1042,"type":42,"tunes":1044},{"text":1043,"level":219},"2. A2A: connect independent agents",{},{"id":357,"data":1046,"type":226,"tunes":1048},{"text":1047},"Agent2Agent (A2A) is designed for communication between independent, potentially opaque agent systems. Its current v1.0 specification focuses on capability discovery, messaging, tasks, artifacts, multimodal content and long-running collaboration without requiring one agent to expose its internal tools, memory or implementation to another.",{},{"id":362,"data":1050,"type":226,"tunes":1052},{"text":1051},"That opacity is the important boundary. The calling agent does not need to know whether the remote agent uses MCP, custom tools, a proprietary planner, another model vendor, or human escalation internally. It needs a contract for discovering capabilities and delegating work.",{},{"id":367,"data":1054,"type":226,"tunes":1056},{"text":1055},"A2A v1.0 also standardizes version negotiation and supports multiple bindings around a common data model. Its published Agent Card mechanism gives clients a standard discovery point for an agent's capabilities, supported protocols, authentication requirements and skills.",{},{"id":372,"data":1058,"type":42,"tunes":1060},{"text":1059,"level":218},"Use A2A when",{},{"id":377,"data":1062,"type":343,"tunes":1069},{"meta":1063,"items":1064,"style":342},{},[1065,1066,1067,1068],"One autonomous agent needs to delegate work to another autonomous agent.","The remote system should remain opaque behind a capability contract.","Tasks may be long-running, asynchronous or require human-in-the-loop interaction.","Agents are built with different frameworks, languages, vendors or organizational ownership.",{},{"id":387,"data":1071,"type":42,"tunes":1073},{"text":1072,"level":219},"MCP vs A2A: vertical integration vs horizontal collaboration",{},{"id":392,"data":1075,"type":421,"tunes":1095},{"rows":1076,"title":1089,"layout":304,"columns":1090},[1077,1080,1083,1086],{"id":396,"label":1078,"values":1079},"Relationship",[399,399,399],{"id":401,"label":1081,"values":1082},"Abstraction",[399,399,399],{"id":405,"label":1084,"values":1085},"Internal opacity",[399,399,399],{"id":409,"label":1087,"values":1088},"Long-running work",[399,399,399],"MCP and A2A solve different interoperability problems",[1091,1093,1094],{"id":415,"label":1092},"Dimension",{"id":418,"label":280},{"id":420,"label":285},{},{"id":424,"data":1097,"type":226,"tunes":1099},{"text":1098},"The A2A project itself now describes the distinction as horizontal versus vertical: MCP connects agents to internal tools and databases, while A2A enables peer-to-peer collaboration across agent systems.",{},{"id":429,"data":1101,"type":42,"tunes":1103},{"text":1102,"level":219},"3. UCP: standardize agentic commerce",{},{"id":434,"data":1105,"type":226,"tunes":1107},{"text":1106},"The Universal Commerce Protocol is not a generic agent protocol. It standardizes commerce journeys between consumer surfaces, merchants and payment providers. Google's implementation already supports capabilities such as cart creation, checkout, fulfillment and order lifecycle through versioned profiles and APIs.",{},{"id":439,"data":1109,"type":226,"tunes":1111},{"text":1110},"A merchant can publish a UCP profile under \u002F.well-known\u002Fucp describing services, protocol versions and capabilities. That discovery pattern matters because an agentic surface should not need a bespoke checkout contract for every merchant.",{},{"id":444,"data":1113,"type":226,"tunes":1115},{"text":1114},"UCP is also intentionally composable. Google's technical overview says it can integrate through APIs, A2A and MCP and is compatible with AP2 for agentic payment authorization.",{},{"id":449,"data":1117,"type":42,"tunes":1119},{"text":1118,"level":218},"Use UCP when",{},{"id":454,"data":1121,"type":343,"tunes":1128},{"meta":1122,"items":1123,"style":342},{},[1124,1125,1126,1127],"The workflow involves merchant products, carts, checkout, fulfillment or order lifecycle.","You are building a merchant surface that should work with agentic shopping experiences.","The integration needs commerce-specific semantics rather than generic tool calls.","You want an interoperable commerce contract that can coexist with MCP, A2A and payment protocols.",{},{"id":464,"data":1130,"type":42,"tunes":1132},{"text":1131,"level":219},"4. AP2: prove that the agent was allowed to spend",{},{"id":469,"data":1134,"type":226,"tunes":1136},{"text":1135},"Agentic commerce introduces a problem that ordinary checkout flows did not have to solve in the same way: an agent may transact when the human is not clicking the final button in real time. The Agent Payments Protocol (AP2) addresses authorization, authenticity and accountability for agent-led payments.",{},{"id":474,"data":1138,"type":226,"tunes":1140},{"text":1139},"Google's 2026 protocol guide describes AP2 through typed mandates that capture user intent, spending constraints and the specific transaction being authorized. AP2 can work as an extension alongside UCP: UCP describes the commerce transaction, while AP2 provides evidence that the agent had authority to perform the payment.",{},{"id":479,"data":1142,"type":226,"tunes":1144},{"text":1143},"This distinction is important. A checkout protocol can tell a merchant what should be purchased. It does not by itself prove who authorized the agent to spend, under what limit, for which merchant, for how long, or whether the final cart remained inside that authority.",{},{"id":484,"data":1146,"type":42,"tunes":1148},{"text":1147,"level":219},"UCP vs AP2: transaction semantics vs authority",{},{"id":489,"data":1150,"type":304,"tunes":1174},{"content":1151,"stretched":43,"withHeadings":14},[1152,1154,1158,1162,1166,1170],[1153,290,295],"Question",[1155,1156,1157],"What is being bought?","Commerce items, cart, checkout and fulfillment semantics","References the authorized transaction context",[1159,1160,1161],"Who may authorize it?","Not the primary protocol responsibility","Explicit agent\u002Fuser authority and mandate model",[1163,1164,1165],"What spending constraints apply?","Commerce flow can contain totals and checkout data","Authorization guardrails and intent limits",[1167,1168,1169],"How is the transaction audited?","Order and commerce lifecycle","Cryptographic \u002F verifiable authorization trail through mandates and receipts",[1171,1172,1173],"Can they work together?","Yes","Yes — AP2 can extend agentic commerce flows",{},{"id":516,"data":1176,"type":42,"tunes":1178},{"text":1177,"level":219},"5. A2UI: let agents describe interfaces without owning your frontend",{},{"id":521,"data":1180,"type":226,"tunes":1182},{"text":1181},"Agent-to-User Interface (A2UI) tackles another boundary: how a remote or local agent communicates a rich interactive interface to a host application. Instead of sending arbitrary HTML, CSS and JavaScript, A2UI uses declarative data that the host renders through its own trusted component catalog.",{},{"id":526,"data":1184,"type":226,"tunes":1186},{"text":1185},"This preserves the host application's design system and security model while still allowing an agent to request dynamic interfaces. A2UI v0.9 specifically emphasizes framework-agnostic UI intent and streaming updates across web, mobile and other clients.",{},{"id":531,"data":1188,"type":226,"tunes":1190},{"text":1189},"Google's later A2UI + MCP Apps work also demonstrates that these UI models are not necessarily mutually exclusive. Declarative native UI and richer embedded application experiences can coexist depending on the task.",{},{"id":536,"data":1192,"type":42,"tunes":1194},{"text":1193,"level":218},"Use A2UI when",{},{"id":541,"data":1196,"type":343,"tunes":1203},{"meta":1197,"items":1198,"style":342},{},[1199,1200,1201,1202],"A remote agent needs to request forms, cards, controls or other interactive UI.","The host should preserve its native components, styling and security boundary.","You do not want remote agents shipping arbitrary executable frontend code.","The same agent-defined UI intent should work across different client frameworks.",{},{"id":551,"data":1205,"type":42,"tunes":1207},{"text":1206,"level":219},"The Protocol Selection Test",{},{"id":556,"data":1209,"type":226,"tunes":1211},{"text":1210},"Do not start from the acronym. Start from the relationship that needs interoperability.",{},{"id":561,"data":1213,"type":587,"tunes":1237},{"steps":1214,"title":1236,"orientation":586},[1215,1218,1221,1224,1227,1230,1233],{"label":1216,"description":1217},"1. Identify the two independent parties","Is this AI-to-tool, agent-to-agent, agent-to-merchant, agent-to-payment authority, or agent-to-user-interface?",{"label":1219,"description":1220},"2. Identify the shared object","Is the contract about a tool call, task, cart, payment mandate, artifact, or UI description?",{"label":1222,"description":1223},"3. Check whether a domain protocol already exists","Prefer commerce or payment semantics when the problem is commerce or authorization instead of encoding everything as generic tools.",{"label":1225,"description":1226},"4. Keep local internals local","Do not expose an entire agent as MCP tools if the remote party only needs an A2A capability, and do not make a remote agent responsible for your UI runtime.",{"label":1228,"description":1229},"5. Compose protocols when the workflow crosses boundaries","One workflow can legitimately cross tool, agent, commerce, payment and UI contracts.",{"label":1231,"description":1232},"6. Version each contract independently","Protocol versions evolve at different speeds; do not tie every integration to one monolithic application version.",{"label":1234,"description":1235},"7. Preserve authorization at every boundary","Interoperability does not replace product permissions, tool authorization, payment authority or data-access policy.","Choose the protocol by the boundary",{},{"id":590,"data":1239,"type":42,"tunes":1241},{"text":1240,"level":219},"A realistic multi-protocol workflow",{},{"id":595,"data":1243,"type":587,"tunes":1264},{"steps":1244,"title":1263,"orientation":586},[1245,1248,1251,1254,1257,1260],{"label":1246,"description":1247},"1. Inspect internal stock with MCP","The purchasing agent calls inventory and forecasting capabilities exposed by internal MCP servers.",{"label":1249,"description":1250},"2. Discover a supplier agent with A2A","The agent reads the supplier's Agent Card and delegates an availability and lead-time task.",{"label":1252,"description":1253},"3. Negotiate the commerce object with UCP","The supplier or merchant surface returns structured cart, checkout and fulfillment information.",{"label":1255,"description":1256},"4. Check spending authority with AP2","The purchase is compared with the user's or organization's signed mandate, merchant constraints and spending limits.",{"label":1258,"description":1259},"5. Ask for approval through A2UI","If human approval is required, the agent sends declarative UI intent and the host renders the approval experience using trusted native components.",{"label":1261,"description":1262},"6. Complete and audit","Commerce state, payment authorization, agent task evidence and application audit records remain traceable across their respective boundaries.","Example: an autonomous procurement workflow",{},{"id":619,"data":1266,"type":42,"tunes":1268},{"text":1267,"level":219},"Why one universal agent protocol is unlikely to replace all of them",{},{"id":624,"data":1270,"type":226,"tunes":1272},{"text":1271},"A universal protocol sounds simpler until it must encode every domain's semantics. Tool discovery, long-running agent collaboration, checkout, payment authorization and native UI all have different lifecycle, security and correctness requirements.",{},{"id":629,"data":1274,"type":226,"tunes":1276},{"text":1275},"The web itself evolved through layered protocols rather than one message format for every problem. The emerging agentic stack appears to be moving in the same direction: common horizontal primitives, specialized domain contracts and explicit discovery\u002Fversioning.",{},{"id":634,"data":1278,"type":226,"tunes":1280},{"text":1279},"The architecture challenge therefore shifts from “which protocol wins?” to how cleanly protocols compose without duplicating identity, authorization, state and audit semantics.",{},{"id":639,"data":1282,"type":42,"tunes":1284},{"text":1283,"level":219},"Protocol composition creates new failure modes",{},{"id":644,"data":1286,"type":304,"tunes":1316},{"content":1287,"stretched":43,"withHeadings":14},[1288,1292,1296,1300,1304,1308,1312],[1289,1290,1291],"Failure mode","What happens","Architecture control",[1293,1294,1295],"Authority leakage","A valid tool or agent capability is treated as permission to perform a business action","Keep product authorization independent from protocol capability discovery",[1297,1298,1299],"Identity mismatch","MCP host identity, A2A agent identity and commerce\u002Fpayment identity refer to different principals","Define explicit principal mapping across boundaries",[1301,1302,1303],"Version drift","One protocol upgrades while dependent adapters assume older semantics","Negotiate and pin protocol versions independently",[1305,1306,1307],"State duplication","The same cart, task or approval state is copied into several protocol layers","Define one authoritative owner per domain object",[1309,1310,1311],"Audit fragmentation","Tool traces, agent tasks, checkout and payment evidence cannot be joined","Carry correlation IDs and stable domain identifiers across protocol boundaries",[1313,1314,1315],"Semantic tunneling","Everything is forced through a generic protocol as opaque JSON","Use domain protocols where their semantics materially improve correctness",{},{"id":677,"data":1318,"type":42,"tunes":1320},{"text":1319,"level":219},"Protocol choice does not replace application architecture",{},{"id":682,"data":1322,"type":226,"tunes":1324},{"text":1323},"Open standards reduce integration coupling, but they do not decide your domain model, authorization policy, source of truth, retry strategy or acceptance criteria. An MCP tool can still expose the wrong capability. An A2A agent can still return a bad artifact. A UCP checkout can still contain stale merchant data. An AP2 mandate can still be misapplied by application logic.",{},{"id":687,"data":1326,"type":226,"tunes":1328},{"text":1327},"Treat protocols as contracts between independently evolving components. Keep domain truth and consequential policy in the application layer that owns them, then use protocols to make the boundaries interoperable.",{},{"id":692,"data":1330,"type":698,"tunes":1335},{"url":1331,"title":1332,"excerpt":1333,"ctaLabel":1334},"https:\u002F\u002Fstajic.de\u002Fblog\u002Fmanaged-agent-harness-vs-self-hosted-agent-loop-what-you-gain-what-you-lose","Managed Agent Harness vs Self-Hosted Agent Loop: What You Gain, What You Lose","Protocol boundaries solve interoperability. Runtime boundaries solve who operates the harness, execution environment and application control plane.","Read the runtime architecture guide",{},{"id":701,"data":1337,"type":42,"tunes":1339},{"text":1338,"level":219},"What would change this answer?",{},{"id":706,"data":1341,"type":226,"tunes":1343},{"text":1342},"The stack changes if protocols converge, one standard formally absorbs another, or vendors standardize a shared identity and authorization layer across several boundaries. UCP already demonstrates composition by supporting APIs, A2A and MCP and by integrating with AP2 rather than replacing them.",{},{"id":711,"data":1345,"type":226,"tunes":1347},{"text":1346},"The answer also changes by application scope. A small internal agent may need only MCP. A multi-company workflow may need A2A. A merchant may need UCP without A2UI. A delegated purchasing agent may need all of them. Use the smallest protocol set that represents the real boundaries without flattening domain semantics.",{},{"id":716,"data":1349,"type":42,"tunes":1351},{"text":1350,"level":219},"Limitations",{},{"id":721,"data":1353,"type":226,"tunes":1355},{"text":1354},"The protocols discussed here are at different maturity levels and have different governance models. A2A has reached a stable v1.0 specification, while other standards continue to evolve rapidly. Ecosystem adoption is also uneven across vendors and frameworks.",{},{"id":726,"data":1357,"type":226,"tunes":1359},{"text":1358},"This article focuses on architecture responsibility rather than implementation completeness. Specific authentication methods, transport bindings, schemas and extension mechanisms must be taken from each protocol's current specification.",{},{"id":731,"data":1361,"type":42,"tunes":1363},{"text":1362,"level":219},"Conclusion",{},{"id":736,"data":1365,"type":226,"tunes":1367},{"text":1366},"MCP, A2A, UCP, AP2 and A2UI make more sense when viewed as protocols for different relationships, not five competing attempts to standardize “agents.”",{},{"id":741,"data":1369,"type":226,"tunes":1371},{"text":1370},"MCP exposes capabilities. A2A coordinates independent agents. UCP gives commerce its own machine-readable contract. AP2 adds verifiable payment authority. A2UI gives agents a safe declarative path into user interfaces. The emerging agentic web is therefore not replacing protocols with AI; it is creating a new protocol stack around AI.",{},{"id":746,"data":1373,"type":42,"tunes":1375},{"text":1374,"level":219},"FAQ",{},{"id":751,"data":1377,"type":751,"tunes":1398},{"items":1378,"title":1397},[1379,1382,1385,1388,1391,1394],{"id":755,"answer":1380,"question":1381},"No. MCP primarily standardizes how AI applications access tools, resources and data. A2A standardizes collaboration between independent agent systems. A remote agent can internally use MCP while exposing an A2A interface.","Is A2A a replacement for MCP?",{"id":759,"answer":1383,"question":1384},"Not generally. UCP provides commerce-specific semantics such as cart, checkout and fulfillment. MCP can still expose merchant tools or data, and UCP is designed to coexist with MCP and A2A.","Is UCP a replacement for MCP in shopping agents?",{"id":763,"answer":1386,"question":1387},"UCP standardizes commerce interactions and transaction lifecycle. AP2 focuses on proving that an agent had authority to perform a payment under defined user or organizational constraints.","What is the difference between UCP and AP2?",{"id":767,"answer":1389,"question":1390},"A2UI lets agents send declarative UI intent to a host application, which renders the experience through trusted native components instead of executing arbitrary remote frontend code.","What problem does A2UI solve?",{"id":771,"answer":1392,"question":1393},"Yes. A workflow can use MCP for internal tools, A2A for remote-agent delegation, UCP for commerce, AP2 for payment authorization and A2UI for human interaction.","Can one agent application use all of these protocols?",{"id":775,"answer":1395,"question":1396},"Start from the interoperability boundary. If the problem is tool access, evaluate MCP. If it is independent-agent collaboration, evaluate A2A. If it is commerce, UCP. If it is delegated payment authority, AP2. If it is portable agent-driven UI, A2UI.","Which protocol should I implement first?","MCP, A2A, UCP, AP2 and A2UI",{},{"id":781,"data":1400,"type":42,"tunes":1402},{"text":1401,"level":219},"Glossary",{},{"id":786,"data":1404,"type":786,"tunes":1420},{"title":1405,"entries":1406},"Key agent-protocol terms",[1407,1409,1411,1413,1415,1417],{"term":280,"anchor":418,"definition":1408},"Model Context Protocol, an open standard for exposing tools, resources and prompts from external systems to compatible AI hosts.",{"term":285,"anchor":420,"definition":1410},"Agent2Agent Protocol, an open standard for discovering and collaborating with independent agent systems through messages, tasks and artifacts.",{"term":290,"anchor":795,"definition":1412},"Universal Commerce Protocol, an open standard for interoperable agentic commerce journeys between consumer surfaces, businesses and payment providers.",{"term":295,"anchor":798,"definition":1414},"Agent Payments Protocol, an open standard for representing and verifying authority, intent and accountability in agent-led payments.",{"term":300,"anchor":801,"definition":1416},"Agent-to-User Interface, a declarative protocol for allowing agents to request UI that is rendered using the host application's trusted components.",{"term":1418,"anchor":805,"definition":1419},"Protocol composition","Using multiple protocols in one workflow, each responsible for a distinct interoperability boundary rather than forcing all semantics through one contract.",{},{"id":809,"data":1422,"type":42,"tunes":1424},{"text":1423,"level":219},"Primary sources and further reading",{},{"id":814,"data":1426,"type":821,"tunes":1431},{"link":816,"meta":1427},{"image":1428,"title":1429,"description":1430},{"url":399},"Google Developers — Developer's Guide to AI Agent Protocols","A practical overview showing MCP, A2A, UCP, AP2, A2UI and related protocols working together in one multi-step agent workflow.",{},{"id":824,"data":1433,"type":821,"tunes":1437},{"link":826,"meta":1434},{"image":1435,"title":829,"description":1436},{"url":399},"Current stable SDK documentation implementing the 2026-07-28 MCP specification and defining tools, resources, prompts and host\u002Fserver integration.",{},{"id":833,"data":1439,"type":821,"tunes":1444},{"link":835,"meta":1440},{"image":1441,"title":1442,"description":1443},{"url":399},"A2A Protocol — v1.0 Specification","Current A2A protocol specification covering Agent Cards, messages, tasks, artifacts, bindings and version negotiation.",{},{"id":842,"data":1446,"type":821,"tunes":1451},{"link":844,"meta":1447},{"image":1448,"title":1449,"description":1450},{"url":399},"A2A — Joining the Agentic AI Foundation","Current project framing of A2A as the horizontal agent-collaboration layer alongside MCP as vertical tool\u002Fdata integration.",{},{"id":851,"data":1453,"type":821,"tunes":1458},{"link":853,"meta":1454},{"image":1455,"title":1456,"description":1457},{"url":399},"Google Developers — Under the Hood: Universal Commerce Protocol","Technical overview of UCP, its commerce primitives and its ability to compose with APIs, A2A, MCP and AP2.",{},{"id":860,"data":1460,"type":821,"tunes":1465},{"link":862,"meta":1461},{"image":1462,"title":1463,"description":1464},{"url":399},"Google Universal Commerce Protocol — UCP Profile","Current versioned profile mechanism for publishing UCP services and merchant commerce capabilities.",{},{"id":869,"data":1467,"type":821,"tunes":1471},{"link":871,"meta":1468},{"image":1469,"title":874,"description":1470},{"url":399},"Announcement and rationale for an open protocol covering authorization, authenticity and accountability in agent-led payments.",{},{"id":878,"data":1473,"type":821,"tunes":1477},{"link":880,"meta":1474},{"image":1475,"title":883,"description":1476},{"url":399},"A2UI's framework-agnostic declarative model for portable agent-driven interfaces rendered by host-native components.",{},{"id":887,"data":1479,"type":821,"tunes":1483},{"link":889,"meta":1480},{"image":1481,"title":892,"description":1482},{"url":399},"How declarative A2UI and richer MCP App experiences can coexist rather than being treated as mutually exclusive UI models.",{},"2.31.6","MCP, A2A, UCP, AP2 and A2UI are often presented as competing agent standards. They mostly solve different interoperability problems. This guide maps each protocol to the boundary it actually standardizes—and shows how they can work together in one production system.",{"lang":7,"title":208,"content":210,"contentJson":1487,"excerpt":896},{"time":212,"blocks":1488,"version":895},[1489,1492,1495,1498,1501,1504,1507,1510,1513,1516,1526,1529,1532,1535,1538,1541,1546,1549,1552,1555,1558,1561,1564,1569,1572,1588,1591,1594,1597,1600,1603,1606,1611,1614,1617,1620,1623,1626,1636,1639,1642,1645,1648,1651,1656,1659,1662,1673,1676,1686,1689,1692,1695,1698,1701,1712,1715,1718,1721,1724,1727,1730,1733,1736,1739,1742,1745,1748,1751,1754,1764,1767,1777,1780,1785,1790,1795,1800,1805,1810,1815,1820],{"id":215,"data":1490,"type":220,"tunes":1491},{"title":217,"maxLevel":218,"minLevel":219},{},{"id":223,"data":1493,"type":226,"tunes":1494},{"text":225},{},{"id":229,"data":1496,"type":234,"tunes":1497},{"body":231,"title":232,"variant":233},{},{"id":237,"data":1499,"type":234,"tunes":1500},{"body":239,"title":240,"variant":241},{},{"id":244,"data":1502,"type":234,"tunes":1503},{"body":246,"title":247,"variant":248},{},{"id":251,"data":1505,"type":42,"tunes":1506},{"text":253,"level":219},{},{"id":256,"data":1508,"type":226,"tunes":1509},{"text":258},{},{"id":261,"data":1511,"type":226,"tunes":1512},{"text":263},{},{"id":266,"data":1514,"type":42,"tunes":1515},{"text":268,"level":219},{},{"id":271,"data":1517,"type":304,"tunes":1525},{"content":1518,"stretched":43,"withHeadings":14},[1519,1520,1521,1522,1523,1524],[275,276,277,278],[280,281,282,283],[285,286,287,288],[290,291,292,293],[295,296,297,298],[300,301,302,303],{},{"id":307,"data":1527,"type":234,"tunes":1528},{"body":309,"title":310,"variant":311},{},{"id":314,"data":1530,"type":42,"tunes":1531},{"text":316,"level":219},{},{"id":319,"data":1533,"type":226,"tunes":1534},{"text":321},{},{"id":324,"data":1536,"type":226,"tunes":1537},{"text":326},{},{"id":329,"data":1539,"type":42,"tunes":1540},{"text":331,"level":218},{},{"id":334,"data":1542,"type":343,"tunes":1545},{"meta":1543,"items":1544,"style":342},{},[338,339,340,341],{},{"id":346,"data":1547,"type":234,"tunes":1548},{"body":348,"title":349,"variant":241},{},{"id":352,"data":1550,"type":42,"tunes":1551},{"text":354,"level":219},{},{"id":357,"data":1553,"type":226,"tunes":1554},{"text":359},{},{"id":362,"data":1556,"type":226,"tunes":1557},{"text":364},{},{"id":367,"data":1559,"type":226,"tunes":1560},{"text":369},{},{"id":372,"data":1562,"type":42,"tunes":1563},{"text":374,"level":218},{},{"id":377,"data":1565,"type":343,"tunes":1568},{"meta":1566,"items":1567,"style":342},{},[381,382,383,384],{},{"id":387,"data":1570,"type":42,"tunes":1571},{"text":389,"level":219},{},{"id":392,"data":1573,"type":421,"tunes":1587},{"rows":1574,"title":412,"layout":304,"columns":1583},[1575,1577,1579,1581],{"id":396,"label":397,"values":1576},[399,399,399],{"id":401,"label":402,"values":1578},[399,399,399],{"id":405,"label":406,"values":1580},[399,399,399],{"id":409,"label":410,"values":1582},[399,399,399],[1584,1585,1586],{"id":415,"label":416},{"id":418,"label":280},{"id":420,"label":285},{},{"id":424,"data":1589,"type":226,"tunes":1590},{"text":426},{},{"id":429,"data":1592,"type":42,"tunes":1593},{"text":431,"level":219},{},{"id":434,"data":1595,"type":226,"tunes":1596},{"text":436},{},{"id":439,"data":1598,"type":226,"tunes":1599},{"text":441},{},{"id":444,"data":1601,"type":226,"tunes":1602},{"text":446},{},{"id":449,"data":1604,"type":42,"tunes":1605},{"text":451,"level":218},{},{"id":454,"data":1607,"type":343,"tunes":1610},{"meta":1608,"items":1609,"style":342},{},[458,459,460,461],{},{"id":464,"data":1612,"type":42,"tunes":1613},{"text":466,"level":219},{},{"id":469,"data":1615,"type":226,"tunes":1616},{"text":471},{},{"id":474,"data":1618,"type":226,"tunes":1619},{"text":476},{},{"id":479,"data":1621,"type":226,"tunes":1622},{"text":481},{},{"id":484,"data":1624,"type":42,"tunes":1625},{"text":486,"level":219},{},{"id":489,"data":1627,"type":304,"tunes":1635},{"content":1628,"stretched":43,"withHeadings":14},[1629,1630,1631,1632,1633,1634],[493,290,295],[495,496,497],[499,500,501],[503,504,505],[507,508,509],[511,512,513],{},{"id":516,"data":1637,"type":42,"tunes":1638},{"text":518,"level":219},{},{"id":521,"data":1640,"type":226,"tunes":1641},{"text":523},{},{"id":526,"data":1643,"type":226,"tunes":1644},{"text":528},{},{"id":531,"data":1646,"type":226,"tunes":1647},{"text":533},{},{"id":536,"data":1649,"type":42,"tunes":1650},{"text":538,"level":218},{},{"id":541,"data":1652,"type":343,"tunes":1655},{"meta":1653,"items":1654,"style":342},{},[545,546,547,548],{},{"id":551,"data":1657,"type":42,"tunes":1658},{"text":553,"level":219},{},{"id":556,"data":1660,"type":226,"tunes":1661},{"text":558},{},{"id":561,"data":1663,"type":587,"tunes":1672},{"steps":1664,"title":585,"orientation":586},[1665,1666,1667,1668,1669,1670,1671],{"label":565,"description":566},{"label":568,"description":569},{"label":571,"description":572},{"label":574,"description":575},{"label":577,"description":578},{"label":580,"description":581},{"label":583,"description":584},{},{"id":590,"data":1674,"type":42,"tunes":1675},{"text":592,"level":219},{},{"id":595,"data":1677,"type":587,"tunes":1685},{"steps":1678,"title":616,"orientation":586},[1679,1680,1681,1682,1683,1684],{"label":599,"description":600},{"label":602,"description":603},{"label":605,"description":606},{"label":608,"description":609},{"label":611,"description":612},{"label":614,"description":615},{},{"id":619,"data":1687,"type":42,"tunes":1688},{"text":621,"level":219},{},{"id":624,"data":1690,"type":226,"tunes":1691},{"text":626},{},{"id":629,"data":1693,"type":226,"tunes":1694},{"text":631},{},{"id":634,"data":1696,"type":226,"tunes":1697},{"text":636},{},{"id":639,"data":1699,"type":42,"tunes":1700},{"text":641,"level":219},{},{"id":644,"data":1702,"type":304,"tunes":1711},{"content":1703,"stretched":43,"withHeadings":14},[1704,1705,1706,1707,1708,1709,1710],[648,649,650],[652,653,654],[656,657,658],[660,661,662],[664,665,666],[668,669,670],[672,673,674],{},{"id":677,"data":1713,"type":42,"tunes":1714},{"text":679,"level":219},{},{"id":682,"data":1716,"type":226,"tunes":1717},{"text":684},{},{"id":687,"data":1719,"type":226,"tunes":1720},{"text":689},{},{"id":692,"data":1722,"type":698,"tunes":1723},{"url":694,"title":695,"excerpt":696,"ctaLabel":697},{},{"id":701,"data":1725,"type":42,"tunes":1726},{"text":703,"level":219},{},{"id":706,"data":1728,"type":226,"tunes":1729},{"text":708},{},{"id":711,"data":1731,"type":226,"tunes":1732},{"text":713},{},{"id":716,"data":1734,"type":42,"tunes":1735},{"text":718,"level":219},{},{"id":721,"data":1737,"type":226,"tunes":1738},{"text":723},{},{"id":726,"data":1740,"type":226,"tunes":1741},{"text":728},{},{"id":731,"data":1743,"type":42,"tunes":1744},{"text":733,"level":219},{},{"id":736,"data":1746,"type":226,"tunes":1747},{"text":738},{},{"id":741,"data":1749,"type":226,"tunes":1750},{"text":743},{},{"id":746,"data":1752,"type":42,"tunes":1753},{"text":748,"level":219},{},{"id":751,"data":1755,"type":751,"tunes":1763},{"items":1756,"title":778},[1757,1758,1759,1760,1761,1762],{"id":755,"answer":756,"question":757},{"id":759,"answer":760,"question":761},{"id":763,"answer":764,"question":765},{"id":767,"answer":768,"question":769},{"id":771,"answer":772,"question":773},{"id":775,"answer":776,"question":777},{},{"id":781,"data":1765,"type":42,"tunes":1766},{"text":783,"level":219},{},{"id":786,"data":1768,"type":786,"tunes":1776},{"title":788,"entries":1769},[1770,1771,1772,1773,1774,1775],{"term":280,"anchor":418,"definition":791},{"term":285,"anchor":420,"definition":793},{"term":290,"anchor":795,"definition":796},{"term":295,"anchor":798,"definition":799},{"term":300,"anchor":801,"definition":802},{"term":804,"anchor":805,"definition":806},{},{"id":809,"data":1778,"type":42,"tunes":1779},{"text":811,"level":219},{},{"id":814,"data":1781,"type":821,"tunes":1784},{"link":816,"meta":1782},{"image":1783,"title":819,"description":820},{"url":399},{},{"id":824,"data":1786,"type":821,"tunes":1789},{"link":826,"meta":1787},{"image":1788,"title":829,"description":830},{"url":399},{},{"id":833,"data":1791,"type":821,"tunes":1794},{"link":835,"meta":1792},{"image":1793,"title":838,"description":839},{"url":399},{},{"id":842,"data":1796,"type":821,"tunes":1799},{"link":844,"meta":1797},{"image":1798,"title":847,"description":848},{"url":399},{},{"id":851,"data":1801,"type":821,"tunes":1804},{"link":853,"meta":1802},{"image":1803,"title":856,"description":857},{"url":399},{},{"id":860,"data":1806,"type":821,"tunes":1809},{"link":862,"meta":1807},{"image":1808,"title":865,"description":866},{"url":399},{},{"id":869,"data":1811,"type":821,"tunes":1814},{"link":871,"meta":1812},{"image":1813,"title":874,"description":875},{"url":399},{},{"id":878,"data":1816,"type":821,"tunes":1819},{"link":880,"meta":1817},{"image":1818,"title":883,"description":884},{"url":399},{},{"id":887,"data":1821,"type":821,"tunes":1824},{"link":889,"meta":1822},{"image":1823,"title":892,"description":893},{"url":399},{},"Post erfolgreich abgerufen",{"items":1827,"source":1877,"manualIds":1878,"manualMatchedIds":1879},[1828,1835,1842,1849,1856,1863,1870],{"id":1829,"slug":1830,"title":1831,"excerpt":1832,"featuredImage":1833,"publishedAt":1834},"467","the-answer-validity-boundary-the-missing-layer-between-relevance-and-reliable-ai-answers","El límite de validez de la respuesta: la capa faltante entre la relevancia y las respuestas fiables de la IA","Una fuente puede ser relevante, autorizada y aun así ser incorrecta para la pregunta que se plantea. La capa que falta es la aplicabilidad: las condiciones bajo las cuales una respuesta es válida y los cambios que obligan a reconsiderarla. Este artículo presenta el Límite de Validez de la Respuesta como un patrón de diseño de fuentes para personas, sistemas de búsqueda con IA y sistemas RAG.","\u002Fuploads\u002F2026\u002F09\u002Fthe-answer-validity-boundary-the-missing-layer-between-relevance-and-reliable-ai-answers-1790272901306-1g5jly.webp","2026-09-24T11:59:00.000Z",{"id":1836,"slug":1837,"title":1838,"excerpt":1839,"featuredImage":1840,"publishedAt":1841},"478","what-is-rag-the-simplest-explanation-of-how-it-works","¿Qué es RAG? La explicación más sencilla de cómo funciona","RAG suena complicado, pero la idea es simple: antes de que una IA responda, primero busca información útil de una fuente de conocimiento y le da esa información al modelo de lenguaje. Esta guía explica RAG, los LLM, el estado, la memoria y las herramientas usando un modelo mental simple.","\u002Fuploads\u002F2026\u002F09\u002Fwhat-is-rag-the-simplest-explanation-of-how-it-works-1790377492124-khjagt.webp","2026-09-25T19:03:00.000Z",{"id":1843,"slug":1844,"title":1845,"excerpt":1846,"featuredImage":1847,"publishedAt":1848},"470","what-should-an-ai-agent-remember-forget-recompute-or-retrieve-again","¿Qué debería recordar, olvidar, recalcular o volver a recuperar un agente de IA?","Los agentes de larga duración no deberían recordarlo todo. Este artículo proporciona un modelo práctico de ciclo de vida para decidir qué pertenece a la memoria duradera, qué se debería recuperar de nuevo, qué es más seguro recalcular y qué debería expirar o ser sustituido.","\u002Fuploads\u002F2026\u002F09\u002Fwhat-should-an-ai-agent-remember-forget-recompute-or-retrieve-again-1790351131087-iehz28.webp","2026-09-25T09:43:00.000Z",{"id":1850,"slug":1851,"title":1852,"excerpt":1853,"featuredImage":1854,"publishedAt":1855},"471","how-to-know-whether-an-ai-agent-actually-used-the-right-evidence","Cómo saber si un agente de IA realmente utilizó la evidencia correcta","Un agente de IA puede citar fuentes y aun así usar la evidencia incorrecta. Este artículo presenta un método práctico para verificar el respaldo de las afirmaciones, la autoridad de la fuente, la aplicabilidad, la procedencia y si la evidencia realmente influyó en la respuesta.","\u002Fuploads\u002F2026\u002F09\u002Fhow-to-know-whether-an-ai-agent-actually-used-the-right-evidence-1790351317188-o5z9ve.webp","2026-09-25T11:47:00.000Z",{"id":1857,"slug":1858,"title":1859,"excerpt":1860,"featuredImage":1861,"publishedAt":1862},"473","openai-agents-api-vs-agents-sdk-vs-responses-api-what-should-you-build-on-in-2026","OpenAI Agents API vs Agents SDK vs Responses API: ¿Sobre qué deberías construir en 2026?","El stack de agentes de OpenAI cambió en septiembre de 2026. Esta guía de arquitectura separa la Agents API, el Agents SDK, la Responses API y el Codex SDK por propiedad del runtime—para que los equipos puedan elegir el límite de control adecuado en lugar de comparar nombres de productos.","\u002Fuploads\u002F2026\u002F09\u002Fopenai-agents-api-vs-agents-sdk-vs-responses-api-what-should-you-build-on-in-2026-1790351846714-zi7lus.webp","2026-09-25T11:56:00.000Z",{"id":1864,"slug":1865,"title":1866,"excerpt":1867,"featuredImage":1868,"publishedAt":1869},"466","the-gpu-is-not-the-product-future-proof-private-ai-architecture","La GPU no es el producto: arquitectura de IA privada a prueba de futuro","La infraestructura de IA privada no debe diseñarse en torno a una sola GPU o un solo modelo. Un enfoque más resiliente combina GPUs de inferencia rápida, sistemas de IA ricos en memoria, nodos de IA física y modelos en la nube frontier opcionales detrás de una capa de enrutamiento consciente de las capacidades.","\u002Fuploads\u002F2026\u002F09\u002Fthe-gpu-is-not-the-product-future-proof-private-ai-architecture-1790140878812-8hsl39.webp","2026-09-23T01:19:00.000Z",{"id":1871,"slug":1872,"title":1873,"excerpt":1874,"featuredImage":1875,"publishedAt":1876},"468","ai-agent-memory-is-not-rag-how-to-separate-memory-retrieval-state-and-context","La memoria del agente de IA no es RAG: cómo separar memoria, recuperación, estado y contexto","La memoria del agente, RAG, el estado y el contexto a menudo se usan como si fueran intercambiables. No lo son. Este modelo práctico de arquitectura separa las cuatro capas, muestra dónde pertenece cada una y explica qué se rompe cuando los sistemas las colapsan en una sola.","\u002Fuploads\u002F2026\u002F09\u002Fai-agent-memory-is-not-rag-how-to-separate-memory-retrieval-state-and-context-1790350560308-np0xy6.webp","2026-09-25T11:34:00.000Z","fallback",[],[]]