Claude 管理代理概述
Claude Managed Agents Overview

原始链接: https://platform.claude.com/docs/en/managed-agents/overview

## Claude 管理代理:摘要 Anthropic 提供两种使用 Claude 的方式:消息 API 用于直接提示,以及 **Claude 管理代理** – 一个用于构建自主代理的完全托管环境。 管理代理提供运行 Claude 的基础设施,具有文件访问、网页浏览和代码执行等功能,*无需*您构建自己的代理循环或沙箱。 关键组件包括 **代理**(定义模型和工具)、**环境**(配置的容器)和 **会话**(运行代理实例)。 **最适合:** 需要云基础设施和有状态会话的长期、异步任务。 它通过处理提示缓存和性能优化来简化开发。 目前处于 Beta 阶段,管理代理提供 Bash、文件操作和网页搜索等内置工具。 访问需要 Claude API 密钥和 `managed-agents-2026-04-01` Beta header。 适用速率限制,并且品牌指南限制将您的产品呈现*为* Anthropic 产品(使用“Claude Agent”或“{您的代理名称} Powered by Claude”代替)。

一个 Hacker News 的讨论围绕着 Anthropic 的“harness”——他们为开发者提供的与 Claude AI 模型交互的工具。虽然承认 Claude 强大的底层 LLM,许多评论者一致认为 harness 本身的代码质量“差到一般”,并且存在一些 bug。 大家的看法是,Anthropic 在人工智能研究方面很强,但没有优先考虑优雅的软件开发。一些人认为他们有意保持控制,而不是开源,这可能是出于对负责任地使用人工智能的担忧。然而,另一些人认为 harness “足够好”以完成任务,并且完善它会减缓 Anthropic 在竞争激烈的市场中的速度。 用户质疑流行度是否等同于质量,尤其是在 Claude 订阅即将出现供应商锁定的情况下,并要求提供更优秀的开源替代方案示例。最终,这场讨论凸显了在快速发展的人工智能领域中,精致与快速开发之间的权衡。
相关文章

原文

Pre-built, configurable agent harness that runs in managed infrastructure. Best for long-running tasks and asynchronous work.

Anthropic offers two ways to build with Claude, each suited to different use cases:

Messages APIClaude Managed Agents
What it isDirect model prompting accessPre-built, configurable agent harness that runs in managed infrastructure
Best forCustom agent loops and fine-grained controlLong-running tasks and asynchronous work
Learn moreMessages API docsClaude Managed Agents docs

Claude Managed Agents provides the harness and infrastructure for running Claude as an autonomous agent. Instead of building your own agent loop, tool execution, and runtime, you get a fully managed environment where Claude can read files, run commands, browse the web, and execute code securely. The harness supports built in prompt caching, compaction, and other performance optimizations for high quality, efficient agent outputs.

Claude Managed Agents is built around four concepts:

ConceptDescription
AgentThe model, system prompt, tools, MCP servers, and skills
EnvironmentA configured container template (packages, network access)
SessionA running agent instance within an environment, performing a specific task and generating outputs
EventsMessages exchanged between your application and the agent (user turns, tool results, status updates)

  1. Define the model, system prompt, tools, MCP servers, and skills. Create the agent once and reference it by ID across sessions.

  2. Configure a cloud container with pre-installed packages (Python, Node.js, Go, etc.), network access rules, and mounted files.

  3. Launch a session that references your agent and environment configuration.

  4. 4

    Send events and stream responses

    Send user messages as events. Claude autonomously executes tools and streams back results via server-sent events (SSE). Event history is persisted server-side and can be fetched in full.

  5. Send additional user events to guide the agent mid-execution, or interrupt it to change direction.

When to use Claude Managed Agents

Claude Managed Agents is best for workloads that need:

  • Long-running execution - Tasks that run for minutes or hours with multiple tool calls
  • Cloud infrastructure - Secure containers with pre-installed packages and network access
  • Minimal infrastructure - No need to build your own agent loop, sandbox, or tool execution layer
  • Stateful sessions - Persistent file systems and conversation history across multiple interactions

Claude Managed Agents gives Claude access to a comprehensive set of built-in tools:

  • Bash - Run shell commands in the container
  • File operations - Read, write, edit, glob, and grep files in the container
  • Web search and fetch - Search the web and retrieve content from URLs
  • MCP servers - Connect to external tool providers

See Tools for the full list and configuration options.

Claude Managed Agents is currently in beta. All Managed Agents endpoints require the managed-agents-2026-04-01 beta header. The SDK sets the beta header automatically. Behaviors may be refined between releases to improve outputs.

To get started, you need:

  1. A Claude API key
  2. The beta header above on all requests
  3. Access to Claude Managed Agents (enabled by default for all API accounts)

Certain features (outcomes, multiagent, and memory) are in research preview. Request access to try them.

Managed Agents endpoints are rate-limited per organization:

OperationLimit
Create endpoints (agents, sessions, environments, etc.)60 requests per minute
Read endpoints (retrieve, list, stream, etc.)600 requests per minute

Organization-level spend limits and tier-based rate limits also apply.

For partners integrating Claude Managed Agents, use of Claude branding is optional. When referencing Claude in your product:

Allowed:

  • "Claude Agent" (preferred for dropdown menus)
  • "Claude" (when within a menu already labeled "Agents")
  • "{YourAgentName} Powered by Claude" (if you have an existing agent name)

Not permitted:

  • "Claude Code" or "Claude Code Agent"
  • "Claude Cowork" or "Claude Cowork Agent"
  • Claude Code-branded ASCII art or visual elements that mimic Claude Code

Your product should maintain its own branding and not appear to be Claude Code, Claude Cowork, or any other Anthropic product. For questions about branding compliance, contact the Anthropic sales team.

联系我们 contact @ memedata.com