I’ve open-sourced DeepMCPAgent, a Python library for building LangChain/LangGraph agents that discover and call tools dynamically via MCP (Model Context Protocol).

Instead of hardcoding tools, the agent connects to MCP servers (HTTP/SSE), fetches JSON-Schema tool specs, converts them into typed LangChain tools, and uses them in a deep agent loop (via deepagents) or a LangGraph ReAct fallback.

Repo: https://github.com/cryxnet/deepmcpagent

PyPI: https://pypi.org/project/deepmcpagent/

Would love feedback on the design and potential use cases. I would love to see what you are going to build!