🚀 DeepMCPAgent 0.4.0 is here!

Hey Product Hunt 👋,

We just shipped DeepMCPAgent 0.4.0, a big step forward in making AI agents more reliable, developer-friendly, and production-ready.

🔑 What’s New in 0.4.0

Better Error Handling → safer defaults and clearer exceptions when MCP tools or servers fail.

Improved Tool Tracing → see exactly which tools were called, with inputs + outputs.

Cleaner CLI → no more flaky auto-gen; we rewrote the CLI docs and polished the experience.

Ruff + Mypy Fixes → linting, typing, and async tests now pass cleanly.

Docs Upgrade → new guides, working examples, and a clean MkDocs setup.

🌐 What is DeepMCPAgent?

DeepMCPAgent is a Python library for building LangChain / LangGraph deep reasoning agents that can dynamically connect to any MCP server.

With built-in discovery and tool recognition, it’s truly plug-and-play: spin up an agent that can talk to multiple servers with just one line of code.

Think of it as agents that don’t just hallucinate answers — they reliably call the tools and servers you control, combining Deep Thinking with ReAct-style reasoning.

💡 Why it matters

Works with any MCP-compliant server

Fully typed, tested, and lightweight

Optional DeepAgents loop for more advanced agent control

Perfect for server orchestration, custom integrations, or private AI workflows

🔧 Example

deepmcpagent run \

--model-id "openai:gpt-4.1" \

--http "name=math url=http://127.0.0.1:8000/mcp transport=http"

→ instantly spin up an agent that can add/multiply via an MCP math server.

📖 Learn More

GitHub: cryxnet/deepmcpagent

PyPI: deepmcpagent

Docs: full guide + API reference

🙌 Thanks for all the feedback — 0.4.0 is about stability and developer experience, but we’re already cooking up a lot of new features for the next release.

Would love your thoughts, questions, and use cases in the comments!