PyMCPX

MCP-compatible LangChain tools for AI agents — calculator, datetime, converter, and more.

Features

• 🔌 MCP-compatible — every tool exposes a stable name and JSON schema consumable by any MCP host
• 🦜 LangChain-native — each tool extends BaseTool with typed args_schema
• 🧱 Pydantic v2 — validated inputs and outputs
• 🧪 Simulation engine — run agent workflows offline without any API keys
• 📦 Service extras — install only what you need: pip install pymcpx[calculator]

Installation

# Core package (no service dependencies)
pip install pymcpx

# Specific service
pip install pymcpx[calculator]
pip install pymcpx[datetime]
pip install pymcpx[converter]

# All services
pip install pymcpx[all]

# Development
pip install pymcpx[dev]

Services

For full details on the tools, schemas, and offline simulation engines of each service, refer to their respective service-level README.md files:

Contributing

Getting Started

git clone https://github.com/your-org/pymcpx.git
cd pymcpx
pip install -e ".[dev]"

Workflow

1. Create a branch: git checkout -b feat/my-feature
2. Make changes following the coding conventions
3. Add / update tests
4. Run tests: pytest
5. Submit a Pull Request

Adding a New Service

Create a new directory under pymcpx/services/<name>/ following the layout of pymcpx/services/ipstack/:

• SimulationEngine/models.py — Pydantic input models
• SimulationEngine/utils.py — business logic, API clients
• tools.py — BaseTool subclasses + MCP_TOOLS list
• SimulationEngine/engine.py — offline simulation engine
• tests/test_<name>.py — unit tests with respx mocks
• README.md — service documentation including:
  • Prerequisites (API key setup)
  • Installation command
  • Tools table with names, classes, descriptions, and input keys
  • Usage examples (individual tools, toolkit, MCP integration)

Then add a re-export shim at pymcpx/<name>.py, register the optional dependency in pyproject.toml, and add the service row to the README Services table.

Pull Request Requirements

• All tests pass
• No new ruff warnings
• New services include all required files (see above)

Continuous Integration

CI (.github/workflows/tests.yml) runs automatically on every push and pull request across Python 3.11 and 3.12 with coverage reporting.

Special Thanks

Special thanks to the Public APIs repository, a manually curated list of public APIs from many domains, which serves as a treasure trove of APIs and inspired our external service integrations.

License

MIT © PyMCPX Contributors