Skip to main content
Get started with the data.gouv.fr MCP Server by connecting your favorite AI chatbot to the hosted endpoint. No installation, API keys, or complex setup required.

​
Connect in 3 steps

1

Choose your AI client

Select an AI chatbot that supports the Model Context Protocol. Popular options include:
  • Claude Desktop - Anthropic’s desktop application
  • ChatGPT - OpenAI’s web interface (Plus, Pro, Team, or Enterprise plans)
  • Gemini CLI - Google’s command-line interface
  • Cursor - AI-powered code editor
  • VS Code - With MCP extension
  • Windsurf - Codeium’s AI editor
  • HuggingChat - Hugging Face’s open chatbot
And 8+ more clients. See the full list with setup instructions for each.
No API key required. The MCP server is read-only and freely accessible.
2

Add the MCP server configuration

Add the hosted endpoint to your client’s MCP configuration file. The exact format depends on your client, but most use a JSON configuration like this:For Claude Desktop (claude_desktop_config.json):
{
  "mcpServers": {
    "datagouv": {
      "command": "npx",
      "args": [
        "mcp-remote",
        "https://mcp.data.gouv.fr/mcp"
      ]
    }
  }
}
For Cursor (MCP settings):
{
  "mcpServers": {
    "datagouv": {
      "url": "https://mcp.data.gouv.fr/mcp",
      "transport": "http"
    }
  }
}
For ChatGPT (web interface):
  1. Go to Settings → Apps and connectors
  2. Enable Developer mode in Advanced settings
  3. Click Add a new connector in the Connectors section
  4. Set the URL to https://mcp.data.gouv.fr/mcp
Configuration file locations vary by platform. On Linux, Claude Desktop uses ~/.config/Claude/claude_desktop_config.json. On macOS, it’s ~/Library/Application Support/Claude/claude_desktop_config.json. See client-specific guides for exact paths.
Using the hosted endpoint at https://mcp.data.gouv.fr/mcp means zero setup. You can also self-host with Docker if needed.
3

Start using the tools

Restart your AI client to load the MCP server. Then ask natural language questions to explore French Open Data:Example questions to try:
  • “What datasets are available about housing prices in Paris?”
  • “Show me the latest population data for Lyon”
  • “Find datasets about renewable energy in France”
  • “Get the visitor statistics for dataset XYZ”
  • “What APIs are available for address geocoding?”
Your AI assistant will automatically use the appropriate MCP tools to search datasets, query data, download files, and retrieve metrics.
The MCP server provides 10 tools for working with datasets, dataservices (external APIs), and metrics. Learn about each tool in the API Reference.

​
What you can do

Once connected, you can use your AI chatbot to:

Search datasets

Find datasets by keywords across 50,000+ resources from data.gouv.fr

Query tabular data

Run queries on CSV and Excel files without downloading them using the Tabular API

Download large files

Parse CSV, JSON, and JSONL files up to 500MB with automatic format detection

Explore APIs

Discover external dataservices and view their OpenAPI specifications

Get usage metrics

Retrieve monthly visits and downloads for any dataset or resource

View metadata

Access detailed information about datasets, resources, organizations, and licenses

​
Example workflow

Here’s a typical conversation flow with your AI assistant:
  1. Search for datasets: “Find datasets about air quality in Paris”
  2. Get dataset details: “Show me more information about the first result”
  3. List resources: “What files are available in this dataset?”
  4. Query the data: “Show me the first 20 rows from the CSV file”
  5. Analyze: “What’s the average PM2.5 level in the data?”
The AI assistant will use multiple MCP tools behind the scenes to fetch the data and answer your questions.
For small datasets, use query_resource_data to preview data via the Tabular API. For large files (>100MB CSV or >12.5MB Excel), use download_and_parse_resource instead.

​
Verify the connection

After adding the configuration and restarting your client, verify the MCP server is connected: In Claude Desktop:
  • Look for the 🔌 plug icon in the chat interface
  • The server name “data.gouv.fr MCP server” should appear in the MCP servers list
In ChatGPT:
  • Go to Settings → Connectors
  • The connector should show as Connected with a green status indicator
In other clients:
  • Check the MCP settings or extensions panel for connection status
  • Some clients show available tools in a sidebar or command palette
If the connection fails, verify that:
  • You’ve entered the exact URL: https://mcp.data.gouv.fr/mcp
  • You’ve restarted your AI client after editing the configuration
  • Your client supports the Model Context Protocol (MCP)

​
Next steps

Client setup guides

Detailed configuration instructions for each supported AI client

Available tools

Learn about all 10 MCP tools with parameters and examples

Query data guide

Best practices for querying datasets and downloading files

Self-host with Docker

Run the MCP server in your own infrastructure

​
Need help?

View on GitHub

Report issues, contribute, or view the source code on GitHub

Build docs developers (and LLMs) love

Get started for freeTalk to us