Google ARD标准落地 · Claude网关自托管

一、 权威必看

EN: The AI agent ecosystem is rapidly moving from experimental prototypes to standardized infrastructure. Google has officially announced the Agentic Resource Discovery (ARD) Specification, an open standard designed to solve the fragmentation problem in publishing, discovering, and verifying AI tools, APIs, and agents. This specification introduces a discovery layer built on catalogs and registries, enabling dynamic capability discovery while leveraging existing protocols such as MCP and OpenAPI for execution. Simultaneously, Google has expanded its Genkit framework by introducing the Agents API, which simplifies the complex plumbing of conversational AI by packaging message history, tool loops, and streaming into a single interface. This allows developers to build agentic full-stack apps with flexible state persistence, supporting advanced workflows like history branching and multi-agent coordination via a unified wire protocol.

中： 人工智能代理（AI Agent）生态正在从实验性原型向标准化基础设施快速演进。Google正式发布了Agent资源发现（ARD）规范，这是一项旨在解决AI工具、API和代理发布、发现及验证碎片化问题的开放标准。该规范引入了基于目录和注册表的发现层，在利用MCP和OpenAPI等现有协议进行执行的同时，实现了动态能力发现。与此同时，Google扩展了其Genkit框架，推出了Agents API，通过将消息历史、工具循环和流式传输打包为单一接口，简化了对话式AI的复杂底层逻辑。这使得开发者能够构建具有灵活状态持久化的全栈代理应用，支持通过统一线协议进行历史分支和多代理协调等高级工作流。

二、 深度与多元

EN: As AI agents become more integrated into enterprise workflows, control and governance have become paramount. AWS and Anthropic have released the Claude apps gateway for AWS, a self-hosted control plane that centralizes identity, policy, telemetry, routing, and spend caps for Claude Code and Claude Desktop. The gateway runs as a single stateless container and routes inference to Amazon Bedrock or Claude Platform on AWS, addressing critical security and cost management concerns for large-scale deployments. In parallel, Google is addressing the linguistic diversity of its global user base with the Gemini Report | Southeast Asia 2026. This report highlights how Gemini is adapting to local languages and cultural contexts in Southeast Asia, demonstrating a strategic shift towards localized AI capabilities that resonate with regional users rather than relying solely on English-centric models.

中： 随着AI代理更深入地融入企业工作流，控制与治理成为核心议题。AWS与Anthropic联合发布了适用于AWS的Claude应用网关，这是一个自托管的控制平面，集中管理Claude Code和Claude Desktop的身份、策略、遥测、路由及支出上限。该网关以单个无状态容器运行，并将推理路由至Amazon Bedrock或AWS上的Claude平台，解决了大规模部署中的关键安全与成本管理问题。并行地，Google通过《Gemini报告 | 东南亚2026》回应了其全球用户群体的语言多样性需求。该报告强调了Gemini如何适应东南亚的当地语言和文化背景，展示了一种向本地化AI能力转变的战略，旨在通过契合区域用户的方式而非仅依赖以英语为中心的模型来建立连接。

三、 科技与财经

EN: The intersection of frontend architecture and agentic AI is yielding new hybrid patterns. Google Developers introduced three architectural patterns to integrate Model Context Protocol (MCP) Apps and Agent-to-User Interface (A2UI), solving the tradeoff between highly custom iframe environments and native, declarative rendering. These patterns allow developers to serve native-feeling UIs directly over MCP servers, embed complex stateful iframe apps securely inside declarative views, or inject generative UI components into legacy systems. On the product front, Gemini is expanding in Chrome to users in the U.K., bringing AI-powered features directly into the browser experience. This move integrates deep search capabilities and contextual assistance into daily browsing, signaling a shift where AI becomes an intrinsic layer of web navigation rather than a separate application.

中： 前端架构与代理式AI的交汇催生了新的混合模式。Google开发者引入了三种架构模式来整合模型上下文协议（MCP）应用与代理到用户界面（A2UI），解决了高度自定义iframe环境与原生声明式渲染之间的权衡问题。这些模式允许开发者通过MCP服务器直接提供具有原生感的UI，安全地将复杂的状态ful iframe应用嵌入声明式视图中，或将生成式UI组件注入遗留系统。在产品方面，Gemini正在向英国用户扩展Chrome中的功能，将AI驱动的功能直接带入浏览器体验。这一举措将深度搜索能力和上下文辅助融入日常浏览中，标志着AI正从独立应用转变为网页导航的内生层。

四、 国际视野

EN: Cloud providers are rapidly enabling AI agents to interact with infrastructure via standardized protocols. Amazon OpenSearch Service now integrates with the Agent Toolkit for AWS, allowing developers to build, manage, and query OpenSearch Service domains and OpenSearch Serverless collections directly from AI coding agents such as Claude Code, Kiro, and Cursor. This integration is powered by the AWS MCP server, which executes AWS API calls on behalf of the agent, paired with a curated amazon-opensearch-service skill that automatically routes natural-language requests to the correct capabilities. This development lowers the barrier for agents to access critical data infrastructure, enabling more autonomous and intelligent search operations without manual API configuration.

中： 云提供商正在通过标准化协议快速使AI代理能够与基础设施进行交互。Amazon OpenSearch Service现已集成AWS Agent Toolkit，允许开发者直接从Claude Code、Kiro和Cursor等AI编码代理构建、管理和查询OpenSearch Service域及OpenSearch Serverless集合。该集成由AWS MCP服务器提供支持，后者代表代理执行AWS API调用，并配合精心策划的amazon-opensearch-service技能，自动将自然语言请求路由至相应功能。这一进展降低了代理访问关键数据基础设施的门槛，使得无需手动配置API即可实现更自主、更智能的搜索操作。

五、 青年与生活

EN: Developer tools are evolving to support the complexity of agentic workflows. Google has released Agent Development Kit (ADK) for Go 2.0, introducing a first-class, graph-based workflow engine to help developers compose complex, multi-agent applications. This update adds built-in primitives for human-in-the-loop (HITL) orchestration, dynamic execution using plain Go code, and automated resilience features like exponential backoff retries. By unifying the execution model, both single-agent applications and intricate graphs now run on the same runtime, simplifying telemetry and management. Additionally, Google introduced a new developer skill for coding agents that automates a five-stage evaluation flywheel: preparing data, running inference, grading with adaptive AutoRaters, analyzing failure clusters, and executing targeted optimizations. This helps developers bridge the gap between prompt tweaks and production stability.

中： 开发者工具正在演进以支持代理工作流的复杂性。Google发布了适用于Go语言的Agent Development Kit (ADK) 2.0，引入了首个一流的基于图的工作流引擎，帮助开发者构建复杂的多代理应用。此次更新增加了内置的人机协作（HITL）编排原语、使用纯Go代码的动态执行功能以及指数退避重试等自动化弹性特性。通过统一执行模型，单代理应用和复杂图表现在都在同一运行时上运行，简化了遥测和管理。此外，Google为编码代理引入了新的开发者技能，自动化了五个阶段的评估飞轮：准备数据、运行推理、使用自适应AutoRaters评分、分析失败集群以及执行针对性优化。这有助于开发者弥合提示词调整与生产稳定性之间的差距。

GitHub also announced that security teams can now create, edit, and manage secret scanning custom patterns via the REST API. This enhancement allows for more granular control over code security, enabling developers to define specific patterns for sensitive data detection without relying solely on default rules.

GitHub还宣布，安全团队现在可以通过REST API创建、编辑和管理秘密扫描自定义模式。这一增强功能允许对代码安全进行更细粒度的控制，使开发者能够定义特定敏感数据检测模式，而不仅仅依赖默认规则。

【21ZHAO 综合判断】

EN: The convergence of standardized discovery (ARD), self-hosted control planes (Claude Gateway), and unified runtime engines (ADK Go 2.0) signals a maturation phase for AI agents. The industry is shifting from building isolated models to orchestrating complex, governed systems.

• Adopt the ARD specification early: As an open standard, it will likely become the de facto protocol for agent interoperability. Developers should structure their APIs and tools to be ARD-compliant to ensure future compatibility.
• Leverage graph-based orchestration: With ADK Go 2.0’s built-in HITL and resilience features, complex multi-agent workflows can now be managed with greater reliability. Use these primitives to handle failure states and human interventions explicitly rather than relying on ad-hoc error handling.

中： 标准化发现（ARD）、自托管控制平面（Claude网关）和统一运行时引擎（ADK Go 2.0）的融合标志着AI代理进入成熟阶段。行业正从构建孤立模型转向编排复杂、受控的系统。

• 早期采用ARD规范：作为开放标准，它很可能成为代理互操作的事实协议。开发者应将其API和工具结构化为符合ARD规范，以确保未来的兼容性。
• 利用基于图的编排：借助ADK Go 2.0内置的HITL和弹性功能，复杂的多代理工作流现在可以更可靠地管理。使用这些原语显式处理故障状态和人工干预，而不是依赖临时的错误处理。

参考来源

• [权威要闻]：Google and Industry Partners Announce Agentic Resource Discovery Specification for AI Agents - 原文链接
• [权威要闻]：Build agentic full-stack apps with Genkit - 原文链接
• [深度解读]：AWS Ships Claude Apps Gateway as Self-Hosted Control Plane for Claude Code and Claude Desktop - 原文链接
• [深度解读]：How Gemini is speaking the language of Southeast Asia - 原文链接
• [科技财经]：A2UI + MCP Apps: Combining the best of declarative and custom agentic UIs - 原文链接
• [科技财经]：We’re expanding Gemini in Chrome to users in the U.K. - 原文链接
• [国际视野]：Amazon OpenSearch Service now supports the Agent Toolkit for AWS with a curated skill - 原文链接
• [国际视野]：Build reliable multi-agent applications with ADK Go 2.0. Discover our new graph-based workflow engine, built-in human-in-the-loop, and dynamic orchestration - 原文链接
• [青年声音]：Driving the Agent Quality Flywheel from Your Coding Agent - 原文链接
• [青年声音]：Manage secret scanning custom patterns via REST API - 原文链接