The Open-Source AI Stack
RSS

The Stack · Agents · Open source

Pydantic AI

Pydantic's MIT agent framework; Python, model-agnostic, type-safe through Pydantic models, native MCP, A2A plus graph multi-agent. 1.0 released September 2025.

Pydantic AI is an MIT-licensed agent framework from the team behind Pydantic, the Python data-validation library. Its distinguishing trait is type-safety: agent inputs and outputs are validated through Pydantic models, which appeals to teams already building on that stack. It is Python-only. It is model-agnostic, with providers installed as explicit extras (Bedrock, Groq, Mistral, and others). Multi-agent work and interoperability are handled through A2A support and a pydantic-graph control-flow library. MCP support is native, with built-in client and server pieces. It reached 1.0 in September 2025 and is actively developed, with a 2.0 line in beta as of May 2026. For open-source AI, Pydantic AI is one of the independent, model-neutral frameworks (alongside LangChain, LlamaIndex, CrewAI, and Mastra) rather than a vendor SDK; there is no house model and no commercial-terms gate on its use.

Sources

Want a follow-up? Ask the chat about Pydantic AI 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