AI Practices AI实践 2d ago Updated 1d ago 更新于 1天前 49

Introducing Claude apps gateway for AWS 推出适用于 AWS 的 Claude 应用网关

Anthropic introduces the Claude Apps Gateway for AWS, a self-hosted control plane designed to centralize access, cost, and policy management for Claude Code and Claude Desktop in enterprise environments. The gateway integrates with standard OpenID Connect (OIDC) providers to issue short-lived tokens, eliminating the need for individual cloud credentials and long-lived secrets on developer machines. Key capabilities include centralized policy enforcement, OpenTelemetry-based telemetry routing, in Anthropic推出Claude Apps Gateway for AWS,为企业部署Claude Code和Claude Desktop提供集中化的访问、成本和策略控制。 网关作为无状态容器运行,支持通过Amazon Bedrock或Claude Platform on AWS进行推理路由,确保数据留在AWS安全边界内。 核心功能包括基于OIDC的身份验证、细粒度的策略管理、OpenTelemetry遥测集成、多区域/账户故障转移以及严格的支出上限控制。 架构设计简化了开发者体验,无需为每个开发人员配置单独的云凭证,通过单一YAML文件和IAM角色即可实现生产级部署。

70
Hot 热度
65
Quality 质量
75
Impact 影响力

Analysis 深度分析

TL;DR

  • Anthropic introduces the Claude Apps Gateway for AWS, a self-hosted control plane designed to centralize access, cost, and policy management for Claude Code and Claude Desktop in enterprise environments.
  • The gateway integrates with standard OpenID Connect (OIDC) providers to issue short-lived tokens, eliminating the need for individual cloud credentials and long-lived secrets on developer machines.
  • Key capabilities include centralized policy enforcement, OpenTelemetry-based telemetry routing, intelligent request routing to Amazon Bedrock or Claude Platform on AWS, and configurable spend caps.
  • Deployment is streamlined via stateless containers on AWS ECS, EKS, or EC2, backed by Amazon RDS for PostgreSQL, allowing organizations to maintain data within their AWS security boundaries.

Why It Matters

This release addresses a critical gap in enterprise AI adoption: governance and operational control. By providing a unified gateway, Anthropic enables IT and security teams to enforce consistent security policies, monitor usage, and manage costs without disrupting developer workflows. This lowers the barrier for large-scale deployment of AI coding assistants by aligning them with existing corporate identity and infrastructure standards.

Technical Details

  • Architecture: The gateway is a stateless service deployed as a container on AWS ECS, EKS, or EC2, sitting behind an internal Application Load Balancer with TLS. It uses Amazon RDS for PostgreSQL to store short-lived sign-in state and rate-limit counters.
  • Authentication & Identity: Integrates with any standards-compliant OIDC provider. Developers authenticate via browser SSO, receiving short-lived tokens that expire based on configured lifetimes (default one hour). Session revocation is immediate upon removal from the IdP.
  • Policy & Routing: Administrators define managed settings (allowed models, tool permissions) in a central YAML configuration. The gateway routes inference requests to upstream providers (Amazon Bedrock or Claude Platform on AWS) using IAM roles, supporting failover across regions or accounts.
  • Telemetry & Cost Control: Usage metrics are stamped by the client and relayed via OpenTelemetry Protocol (OTLP) to collectors like Amazon CloudWatch. The system supports granular spend caps (daily, weekly, monthly) per user, group, or organization, blocking requests when limits are exceeded.

Industry Insight

  • Security First Approach: Enterprises should prioritize deploying this gateway to eliminate the risk of leaked API keys and ensure that sensitive code interactions remain within defined security perimeters.
  • Cost Optimization: Implementing strict spend caps and granular telemetry monitoring will be essential for managing the variable costs associated with LLM usage at scale, preventing budget overruns.
  • Operational Efficiency: Leveraging existing OIDC and device management tools for onboarding/offboarding reduces administrative overhead, making AI tool integration smoother and more scalable across large development teams.

TL;DR

  • Anthropic推出Claude Apps Gateway for AWS,为企业部署Claude Code和Claude Desktop提供集中化的访问、成本和策略控制。
  • 网关作为无状态容器运行,支持通过Amazon Bedrock或Claude Platform on AWS进行推理路由,确保数据留在AWS安全边界内。
  • 核心功能包括基于OIDC的身份验证、细粒度的策略管理、OpenTelemetry遥测集成、多区域/账户故障转移以及严格的支出上限控制。
  • 架构设计简化了开发者体验,无需为每个开发人员配置单独的云凭证,通过单一YAML文件和IAM角色即可实现生产级部署。

为什么值得看

对于希望在企业环境中规模化使用AI编码助手(如Claude Code)的组织而言,该网关解决了分散管理带来的治理、安全和成本失控难题。它提供了统一的控制平面,使IT和安全团队能够像管理传统云服务一样管理LLM应用,确保合规性并优化支出。

技术解析

  • 架构与部署:网关以无状态容器形式运行在AWS基础设施上(支持ECS、EKS或EC2),后端连接Amazon RDS for PostgreSQL存储短期登录状态和速率限制计数器。前端通过内部Application Load Balancer暴露,并使用AWS Certificate Manager管理TLS证书。
  • 身份与策略管理:集成标准OIDC身份提供商,通过浏览器SSO进行单点登录。管理员可集中定义策略(如允许使用的模型、工具权限、默认设置),并按IdP组作用域实施,客户端在登录时自动接收并应用这些设置。
  • 遥测与路由:客户端为每个请求标记使用指标,网关通过OpenTelemetry Protocol (OTLP)将遥测数据转发至配置的收集器(如CloudWatch或Prometheus)。网关持有上游凭证,代表开发者向Amazon Bedrock或Claude Platform on AWS路由推理请求,支持跨Region或多账户的故障转移。
  • 成本控制与安全:实施每日、每周和每月的支出上限,超出限额后自动阻止请求。不依赖长期存在的秘密凭证,会话令牌有效期短(默认一小时),用户从IdP移除后会话随即过期,确保无敏感数据残留于开发机。

行业启示

  • LLM运维标准化:随着AI代理和编码助手的普及,企业需要专门的“LLM网关”层来管理身份、计费和安全策略,这将成为企业AI基础设施的标准组件。
  • 数据主权与合规优先:通过支持Amazon Bedrock等托管服务,Anthropic强调了在公有云环境中保持数据隐私和控制权的重要性,满足了严格监管行业的需求。
  • 降低采用门槛:通过抽象底层API复杂性并提供集中式管理,此类工具降低了企业大规模部署AI工具的行政和技术负担,加速了AI在开发工作流中的渗透。

Disclaimer: The above content is generated by AI and is for reference only. 免责声明:以上内容由 AI 生成,仅供参考。

Claude Claude Deployment 部署 Security 安全