evidoria

← Back to browse

Good practice Imported

data.gouv.fr's MCP Server — Connecting AI Assistants to France's Full Open-Data Catalogue

France · Paris · See the France profile · See the Paris profile

Evidence: Descriptive / self-reported Top 75% 47/100 · Ask Evidence Copilot about this practice

In February 2026 Etalab launched a public MCP server exposing data.gouv.fr's full catalogue — ~74,000 datasets/APIs, 390,000 files — to AI assistants like Claude and ChatGPT in read-only mode, an early such connector for a European open-data portal.

74,000 datasets/APIs
data.gouv.fr catalogue size (datasets and APIs) (as of Jul 2026)
390,000 files
Files represented in the catalogue (as of Jul 2026)
data.gouv.fr's MCP Server — Connecting AI Assistants to France's Full Open-Data Catalogue

Details

Maturity
Pilot
Promoter
Etalab (Direction interministérielle du numérique, DINUM)
Period
February 2026–ongoing (experimental)
Region (NUTS)
FR10
Keywords
open data, digital government, AI infrastructure, interoperability

Context

In February 2026, Etalab — the French digital administration's open-data unit, part of DINUM — launched a Model Context Protocol (MCP) server for data.gouv.fr, the national open-data platform, making it one of the first European government open-data portals to connect to the MCP standard Anthropic introduced in late 2024 for letting AI assistants query external tools and data sources directly.

Activities

The server, hosted publicly at mcp.data.gouv.fr, lets AI chatbots such as Claude, ChatGPT, Mistral's Le Chat, and Gemini search and explore data.gouv.fr's catalogue — about 74,000 datasets and APIs, representing roughly 390,000 files as of July 2026 — through natural-language queries instead of manual browsing, with access strictly read-only (exploration only, no modification of public data).

Results

The government's own announcement frames the launch as an "expérimentation" and publishes a public retour d'expérience (retex) addressing open questions about AI access to public data. Independent coverage (Le Monde Informatique) likewise describes it as an experiment rather than a finished, evaluated service, and no usage or query-volume figures had been published as of August 2026.

Implementation
Indicative cost
Low (< €50k)
Time to results
Short (< 1 year)
Staffing & skills
Etalab (part of DINUM, the French interministerial digital directorate) built and hosts the server

Conditions for success

  • Built on the Model Context Protocol (MCP) standard, enabling multiple AI assistants (Claude, ChatGPT, Mistral's Le Chat, Gemini) to connect without bespoke integrations
  • Access kept strictly read-only, limiting risk while the government gathers experience
  • A public retour d'expérience (retex) documents open questions about AI access to public data

Common failure modes

  • The government's own announcement frames the launch as an 'expérimentation' rather than a finished service
  • No usage or query-volume figures had been published as of August 2026

Where it fits

Governance type
national digital administration unit (interministerial)
Scale
national
Income level
high-income (EU member state)

Commonly funded by

National / regional programmes Digital Europe Programme

Indicative funding routes for practices of this type — always check each programme's current calls and eligibility rules.

Do you run this practice? Claim it — verified implementers get a public contact pathway and can propose corrections.

Data sources

Where this practice's information was retrieved from, and when.

Attachments

Similar practices you may find useful