OpenDataMCP

OpenDataMCP provides an Open Data focused Model Context Protocol (MCP) server implementation built in Python. It offers a CLI-based workflow to access, query, and publish open datasets so you can surface data to LLMs in a structured, governed manner. The repository ships with a provider framework and templates to add new data sources as Tools or Resources, enabling active querying, parameter-based filtering, and data publishing templates. You can run the MCP server to expose your data through the provided endpoints and use the experimental client to validate provider endpoints within an MCP-enabled workflow.

To use it, install the required tooling (UV for the CLI workflow, and Python dependencies for the MCP server), then start the server via the Python module entry point. The project emphasizes a provider-based approach: add a module under src/odmcp/providers, implement the Tool/Resource interfaces, and test with the included client utilities. The OpenDataMCP CLI (uvx odmcp) drives setup and management of provider MCP servers in your Claude or LLM workflow, enabling you to connect LLMs to open data sources and publish datasets through templates and guidelines from the project.

Prerequisites

• Python 3.9+ and pip
• uv (UV Package Manager) for CLI workflow management (macOS/Windows/Linux)
• Git

Installation steps

1. Install UV (prerequisite for the CLI tooling)

• macOS: brew install uv
• Windows (PowerShell): powershell -ExecutionPolicy ByPass -c "irm https://astral.sh/uv/install.ps1 | iex"
• Linux/WSL: curl -LsSf https://astral.sh/uv/install.sh | sh

2. Install Python package odmcp from PyPI (or install from source if contributing)

• Python package (recommended): python -m pip install --upgrade pip pip install odmcp

3. Prepare repository and dependencies (optional for development):

• git clone https://github.com/OpenDataMCP/OpenDataMCP.git
• cd OpenDataMCP
• python -m pip install -e .[dev] # if developing with extras

4. Run the MCP server (as described by the project):

• python -m odmcp

5. Validate the setup with the experimental client as described in the docs (uv run ...)

Notes

• Ensure you have network access to fetch datasets or remote providers configured by your MCP setup.
• If running CLI tooling with Claude Desktop integration, follow the provider setup steps in the repository to configure specific MCP providers.

Environment and configuration tips:

• The provider modules live under src/odmcp/providers/. Create new modules following the naming convention {country_code}_{organization}.py (e.g., ch_sbb.py) and implement the required Tools/Resources with clear input/output schemas using Pydantic.
• Use the experimental UV client to validate endpoints: uv run src/odmcp/providers/client.py
• For publishing data, UV-based workflows and pre-commit hooks help maintain code quality; install pre-commit and run uv sync as described in the docs.
• If you encounter issues with CI or packaging, ensure Python dependencies match the project’s pinned versions and that you’re using a compatible UV version for your OS.