ckan

CKAN MCP Server exposes a conversational interface to any CKAN open data portal. Once connected, your AI tool can search datasets, browse organizations, read dataset metadata, and query tabular data through natural language prompts. The server implements MCP-compatible endpoints so clients can issue requests and receive structured results, enabling workflows like data discovery, metadata exploration, and data retrieval without requiring intimate CKAN API knowledge. You can use the hosted endpoint or run the server locally to integrate with your preferred toolchains (ChatGPT, Claude, Gemini, VS Code, Codex, etc.). The repository provides example configurations for popular clients, including how to point tools at the hosted URL (https://ckan-mcp-server.andy-pr.workers.dev/mcp) or how to install and run the server locally with npm/npx. Tools listed include ckan_package_search for searching CKAN datasets, with additional capabilities typically exposed by MCP adapters such as dataset retrieval, organization exploration, and metadata reads depending on client support.

Prerequisites:

• Node.js and npm installed on your system (Node.js 14+ recommended)
• Optional: Docker if you prefer containerized usage

Option A — Run locally (npm/npx)

1. Install globally (optional): npm install -g @aborruso/ckan-mcp-server
2. Run the server with npx (no global install required): npx @aborruso/ckan-mcp-server@latest
3. The server will start and expose the MCP endpoint at a default port (check console output). You can configure clients to connect to http://localhost:3000/mcp or the port shown by the server.

Option B — Install via Docker

1. Ensure Docker is installed and running.
2. Follow the project/docker/README.md guidance to build and run the container: git clone https://github.com/ondata/ckan-mcp-server.git cd ckan-mcp-server docker compose up --build -d
3. The MCP endpoint will be available at http://localhost:3000/mcp as described in the Docker guide.

Tips and notes:

• Hosted endpoint is available for quick testing without installation: https://ckan-mcp-server.andy-pr.workers.dev/mcp
• If you plan to run locally, using npx/@aborruso/ckan-mcp-server@latest avoids managing versions manually.
• Popular clients (ChatGPT, Claude, Gemini, VS Code, Codex) include example configurations in the README for quick setup.
• The server adheres to public CKAN API principles and DCAT metadata, enabling rich dataset discovery and metadata access via MCP.
• If you encounter rate limits or quotas in hosted mode, consider running locally for unlimited usage as described in the README.