PolyMCP is a universal MCP toolkit and agent framework for Python and TypeScript. It enables you to expose Python or TypeScript tools as MCP servers, connect them via HTTP, stdio, or in-process transports, and orchestrate them with agents. You can create MCP HTTP servers from plain Python functions, run tools over HTTP using the provided expose_tools_http helper, and compose autonomous workflows with the PolyClaw system. The Inspector app provides a fast way to test servers, prompts, and UI resources, while the Agents ecosystem lets you orchestrate multiple servers and even combine Python and TypeScript components in a single workflow. Typical usage includes defining tools in Python, exposing them over HTTP, and then building higher-level agents that call those tools as needed.

To use PolyMCP, install the Python package, start a server (for example via uvicorn for an HTTP server), and connect clients or inspectors to the server endpoint. You can also leverage the CLI to initialize projects, add MCP servers, and run agents. The TypeScript side (polymcp-ts) offers a parallel implementation and runnable use cases, allowing you to build TypeScript MCP services or integrate with the Inspector and MCP Apps SDK. Tools can be exposed with HTTP transports or other MCP transports, and agents can orchestrate tool calls, data processing, and prompts across multiple servers.

Prerequisites:

• Python 3.8+ (recommended 3.8 or newer)
• pip (version compatible with your Python)

1. Create a virtual environment (optional but recommended): python -m venv .venv source .venv/bin/activate # on Unix or macOS ..venv\Scripts\activate # on Windows

2. Install PolyMCP from PyPI: pip install polymcp

3. (Optional) Install typing and dev tools for development: pip install -e .[dev] # if you are working from a checkout

4. If you plan to use TypeScript tooling as well, install Node.js 18+ and the polymcp-ts package:

   in a separate workspace

   cd polymcp-ts npm install npm run build

5. Run a sample HTTP server (Python):

   Example using a simple HTTP server exposed from functions

   uvicorn server:app --host 0.0.0.0 --port 8000

6. Verify the server is running by hitting the endpoint (e.g., http://localhost:8000/mcp) and using the Inspector or CLI tooling to interact with it.

Notes and tips:

• PolyMCP supports multiple transports (HTTP, stdio, in-process). Choose the transport that fits your deployment scenario.
• The Inspector provides a quick playground to test servers, tools, prompts, and UI resources without building a full UI. Use it to validate tool calls and responses.
• Skills and agents: PolyMCP includes an agent framework for orchestrating multiple servers. You can integrate OpenAI or other providers as LLM backends for planning and tool selection.
• Environment variables: you may want to configure API keys (e.g., OPENAI_API_KEY), host/port bindings, and Docker-related settings if you experiment with PolyClaw.
• For TypeScript users, polymcp-ts offers a parallel runtime in TypeScript; you can build cross-language workflows by combining Python and TS servers.
• If you encounter issues with server startup, ensure dependencies are installed, your Python path is correct, and the port you choose is not in use. Check the Inspector logs and the CLI help messages for troubleshooting guidance.