The Open-Source AI Stack
RSS

The Stack · Agents · Open source

Claude Agent SDK

Anthropic's agent SDK extracted from Claude Code; the package is MIT but its use is governed by Anthropic's Commercial Terms, and it is Claude-only. Native plus in-process MCP, subagents. Renamed from the Claude Code SDK September 2025.

The Claude Agent SDK is the harness that runs Claude Code, made available as a library in Python and TypeScript. It provides the agent loop, context management, tool permissioning, subagents, and MCP integration that the Claude Code product uses. It was renamed from the "Claude Code SDK" on 2025-09-29 to signal use beyond coding. It is the one clear outlier in the 2026 agent-SDK comparison, on two axes. License: the package source is MIT, but the README states that use of the SDK is governed by Anthropic's Commercial Terms of Service, so the open-source license on the code does not make its use unrestricted. Model coupling: it is Claude-only, with no provider-agnostic path. Serving can run through the Anthropic API, Amazon Bedrock, Google Vertex AI, or Microsoft Foundry on Azure, but the model is always Claude. For open-source AI this is the example that separates open code from open use: you can read and fork the SDK, but you cannot point it at an open-weights model or run it outside Anthropic's commercial terms. MCP support is native and includes in-process SDK MCP servers for custom tools. It is pre-1.0 in versioning (0.2.x as of May 2026) but powers a widely deployed product.

Sources

Want a follow-up? Ask the chat about Claude Agent SDK in context. It will compare to siblings at the same layer and ground every claim in the wiki.

Other projects at the Agents layer

19 siblings · ordered open first