Data Commons MCP

🚀 EOSC Data Commons MCP server

A Model Context Protocol (MCP) server that exposes an HTTP POST endpoint for accessing data from various open - access data publishers. It is developed for the EOSC Data Commons project.

This server uses a search API and a Large Language Model (LLM) to assist users in finding the datasets and tools they need.

✨ Features

🧩 Endpoints

The HTTP API consists of 2 main endpoints:

• /mcp: The MCP server searches for relevant data to answer user questions using the EOSC Data Commons search API.

  • It uses rmcp with Streamable HTTP transport.
  • Available tools:
    • [x] Search datasets
    • [ ] Search tools
    • [ ] Search citations related to datasets or tools

• /search: A simple HTTP POST endpoint (JSON) for querying the MCP server via an LLM provider.

  • It uses axum, utoipa for OpenAPI spec generation, and llm to interact with LLM providers (e.g., Mistral, OpenAI).
  • It returns a streaming response: first the requested tool call, then the tool call results, and finally the search results.

🚀 Quick Start

🛠️ Development

⚠️ Important Note

Requirements:

• Rust
• API key for an LLM provider: Mistral.ai or OpenAI. You can use the free tier; you just need to log in.

💡 Usage Tip

It is recommended to use the VSCode extension: rust-analyzer

📥 Install dev dependencies

rustup update
cargo install cargo-release cargo-deny cargo-watch git-cliff

Create a .cargo/config.toml file with your Mistral API key or OpenAI API key:

[env]
MISTRAL_API_KEY = "YOUR_API_KEY"
OPENAI_API_KEY = "YOUR_API_KEY"

⚡️ Start dev server

Start the MCP server in development mode at http://localhost:8000/mcp, with OpenAPI UI at http://localhost:8000/docs.

cargo run

Run and reload the server when the code changes:

cargo watch -x run

Example curl request:

curl -X POST http://localhost:8000/search -H "Content-Type: application/json" -H "Authorization: SECRET_KEY" -d '{"messages": [{"role": "user", "content": "data about insulin in EU"}], "model": "mistral-small-latest", "stream": true}'

🔌 Connect MCP client

Follow the instructions of your client and use the /mcp URL of your deployed server (e.g., http://localhost:8000/mcp).

🐙 VSCode GitHub Copilot

Add a new MCP server through the VSCode UI:

• Open the Command Palette (ctrl+shift+p or cmd+shift+p).
• Search for MCP: Add Server....
• Choose HTTP and provide the MCP server URL http://localhost:8000/mcp.

Your VSCode mcp.json should look like this:

{
    "servers": {
        "data-commons-mcp-server": {
            "url": "http://localhost:8000/mcp",
            "type": "http"
        }
    },
    "inputs": []
}

📦 Build for production

Build the binary in target/release/:

cargo build --release

⚠️ Important Note

Start the server with:

./target/release/data-commons-mcp

🐳 Deploy with Docker

Create a keys.env file with the API keys:

MISTRAL_API_KEY=YOUR_API_KEY
SEARCH_API_KEY=SECRET_KEY_YOU_CAN_USE_IN_FRONTEND_TO_AVOID_SPAM

💡 Usage Tip

SEARCH_API_KEY can be used to add a layer of protection against bots that might spam the LLM. If not provided, no API key will be needed to query the API.

Build and deploy the service:

docker compose up

🧼 Format & lint

Automatically format the codebase using rustfmt:

cargo fmt

Lint the code with clippy:

cargo clippy --all

Automatically apply possible fixes:

cargo fix

⛓️ Check supply chain

Check the dependency supply chain for licenses (only accept dependencies with OSI or FSF approved licenses) and vulnerabilities (CVE advisories):

cargo deny check

Update dependencies in Cargo.lock:

cargo update

🏷️ Release

Dry run:

cargo release patch

Or minor / major

Create a release:

cargo release patch --execute