I built PolyMCP, a framework to make building MCP (Model Context Protocol) servers and integrating agents much easier.
Working with MCP and agent tooling often comes with repetitive or frustrating challenges:
•Exposing Python functions as discoverable tools can be tedious.
•Orchestrating multiple MCP servers usually requires a lot of glue code.
•Debugging multi-server setups is tricky without good visibility into calls, inputs, and outputs.
•Integrating agents with LLMs to automatically discover and call tools is still immature in most setups.
PolyMCP addresses these pain points:
•Flexible tool exposure – Expose Python functions as MCP tools with minimal boilerplate. Supports HTTP, in-process, or stdio execution.
•Real-time Inspector – Monitor tool calls live, inspect inputs/outputs, and test interactions interactively.
•Built-in agent support – Agents can automatically discover and invoke tools with multiple LLM providers.
•CLI and workflow tooling – Quickly scaffold, test, and run MCP projects without tedious setup.
The goal is to make MCP server development and multi-tool agent orchestration faster and less painful.
Check it out here: https://github.com/poly-mcp/Polymcp
I’d love to hear feedback, ideas, or suggestions!