data.gouv.fr MCP — Use Cases, Install & Live Demo

Name: data.gouv.fr MCP Server
Author: datagouv

Why use it

Key features

Full-text search across tens of thousands of public datasets
Metadata (license, publisher, update frequency, tags) for each dataset
Resource URLs (CSV/JSON/XLSX) for direct download
No API key — entirely public data
Coverage: demographics, transport, environment, health, elections, INSEE etc.

Live Demo

What it looks like in practice

datagouv.replay ▶ ready

0/0

Install

Pick your client

~/Library/Application Support/Claude/claude_desktop_config.json · Windows: %APPDATA%\Claude\claude_desktop_config.json

{
  "mcpServers": {
    "datagouv": {
      "command": "uvx",
      "args": [
        "datagouv-mcp"
      ]
    }
  }
}

Open Claude Desktop → Settings → Developer → Edit Config. Restart after saving.

~/.cursor/mcp.json · .cursor/mcp.json

{
  "mcpServers": {
    "datagouv": {
      "command": "uvx",
      "args": [
        "datagouv-mcp"
      ]
    }
  }
}

Cursor uses the same mcpServers schema as Claude Desktop. Project config wins over global.

VS Code → Cline → MCP Servers → Edit

{
  "mcpServers": {
    "datagouv": {
      "command": "uvx",
      "args": [
        "datagouv-mcp"
      ]
    }
  }
}

Click the MCP Servers icon in the Cline sidebar, then "Edit Configuration".

~/.codeium/windsurf/mcp_config.json

{
  "mcpServers": {
    "datagouv": {
      "command": "uvx",
      "args": [
        "datagouv-mcp"
      ]
    }
  }
}

Same shape as Claude Desktop. Restart Windsurf to pick up changes.

~/.continue/config.json

{
  "mcpServers": [
    {
      "name": "datagouv",
      "command": "uvx",
      "args": [
        "datagouv-mcp"
      ]
    }
  ]
}

Continue uses an array of server objects rather than a map.

~/.config/zed/settings.json

{
  "context_servers": {
    "datagouv": {
      "command": {
        "path": "uvx",
        "args": [
          "datagouv-mcp"
        ]
      }
    }
  }
}

Add to context_servers. Zed hot-reloads on save.

claude mcp add datagouv -- uvx datagouv-mcp

One-liner. Verify with claude mcp list. Remove with claude mcp remove.

Use Cases

Real-world ways to use data.gouv.fr

Find a French public dataset by topic

👤 Journalists, researchers, civic-tech devs ⏱ ~10 min beginner

When to use: You need 'unemployment by département, monthly' or similar and don't know where it lives.

Flow

Search by keywords

Search data.gouv.fr for 'chômage département mensuel'. Return top 10 datasets with title, publisher, last update.✓ Copied

→ Relevant dataset list
Inspect the most promising

For the top 3, fetch full metadata: license, update frequency, resources, description. Which is the most current + official?✓ Copied

→ Comparison summary
Get download URLs

Return the CSV resource URL(s) for the chosen dataset. I'll download it separately.✓ Copied

→ Direct download URL(s)

Outcome: A shortlist of authoritative datasets with direct download links.

Pitfalls

Datasets sometimes have broken resource URLs — Check the 'last_modified' metadata; older + unmaintained = more likely broken. Prefer INSEE-published if available.

Combine with: filesystem

Catalog all datasets published by a specific ministry or agency

👤 Researchers mapping data availability ⏱ ~15 min intermediate

When to use: You want to know everything INSEE or Ministère de la Transition écologique publishes.

Flow

Look up the organization

Find the data.gouv organization page for 'INSEE'. Return its slug.✓ Copied

→ Canonical org id
List all datasets

List every dataset published by that org. Page through all results. Return title, id, last update, resource count.✓ Copied

→ Full catalog
Categorize

Group by theme (population, economy, territory, etc.). Count datasets per theme.✓ Copied

→ Thematic histogram

Outcome: A structured inventory of a French public org's open data.

Pitfalls

Large orgs have 1000+ datasets — pagination matters — Always page through all results; don't trust the first page as complete

Combine with: filesystem

Watch for newly published or updated datasets on a topic

👤 Civic tech / policy analysts ⏱ ~15 min intermediate

When to use: You want to be alerted when new COVID/election/housing datasets appear.

Flow

Baseline search today

Search for '<topic>' and save the top 50 dataset ids + last_modified dates as /data/datagouv-<topic>-baseline.json.✓ Copied

→ Baseline saved
Weekly rerun

Rerun the search. Compare against baseline. Return (a) new datasets not in baseline, (b) datasets whose last_modified increased.✓ Copied

→ Delta report
Surface the interesting ones

Filter the delta by official publishers (ministries, INSEE, agencies). Produce a 1-line summary per new/updated dataset.✓ Copied

→ Curated digest

Outcome: A weekly alert on authoritative new French data in your topic.

Pitfalls

Aggregators re-publish old data with new last_modified — noise — Prefer datasets whose publisher is an official org, not a re-publishing community

Combine with: filesystem

Combinations

Pair with other MCPs for X10 leverage

datagouv + filesystem

Download and archive datasets locally

Search 'recensement population' on data.gouv, pick the INSEE one, download CSV, save to /data/insee/.✓ Copied

datagouv + postgres

Ingest a French open dataset into Postgres for analysis

Download the latest unemployment dataset, infer schema, CREATE TABLE and COPY into Postgres.✓ Copied

datagouv + antv-chart

Quick chart from a French open dataset

Download dataset X, filter to 2020-2024, render as a line chart per région.✓ Copied

Tools

What this MCP exposes

Tool	Inputs	When to call	Cost
search_datasets	q: str, page?, page_size?	Keyword search over datasets	free
get_dataset	dataset_id_or_slug	Full metadata + resources for a specific dataset	free
search_organizations	q: str	Find a ministry/agency's canonical id	free
list_organization_datasets	org_id, page?	Catalog a publisher's full output	free

Cost & Limits

What this costs to run

API quota: Public API; fair-use throttling (few req/sec is safe)
Tokens per call: Dataset metadata: 500–2000 tokens depending on field count
Monetary: Free
Tip: Cache dataset ids locally; metadata rarely changes week-over-week.

Security

Permissions, secrets, blast radius

Credential storage: None needed

Data egress: All calls to www.data.gouv.fr and resource hosts

Data is public but licenses vary (Licence Ouverte most common); always check the dataset's license before redistribution.

Troubleshooting

Common errors and fixes

404 on get_dataset

Dataset id changed or was delisted. Use search instead of a cached id.

Rate limited

Space out calls; <5 req/sec is safe. No official limit but fair-use enforced.

Resource URL returns 403/404

Some datasets have broken links; prefer the most recently updated resource or the official publisher's own page.

Alternatives

data.gouv.fr vs others

Alternative	When to use it instead	Tradeoff
data.europa.eu MCP	You need pan-EU data, not just France	Broader but less French-specific depth
INSEE API MCP	You only need official INSEE stats with clean API access	Narrower scope; French statistical office only

More

Resources

📖 Read the official README on GitHub

🐙 Browse open issues

🔍 Browse all 400+ MCP servers and Skills