Skill Detail

Find and inspect French public datasets through data.gouv.fr MCP

Search data.gouv.fr, check dataset provenance and resource metadata, then preview supported tabular rows and return a cited research handoff.

Research & ScrapingMCP
Research & Scraping MCP Security Reviewed
⭐ 1.6k GitHub stars
COPY SKILL INSTRUCTIONS (OPTIONAL)
npx skills add agentskillexchange/skills --skill find-inspect-french-public-datasets-datagouv-mcp Copy
Uses the third-party skills CLI, not an ASE-owned installer. Check your agent’s compatibility. This copies instructions; complete the upstream tool setup below separately.
At a glance
Tools required
MCP client supporting Streamable HTTP; network access to mcp.data.gouv.fr
Install & setup
Configure a remote Streamable HTTP MCP connection to https://mcp.data.gouv.fr/mcp using your client configuration. The official README states no API key is required for its read-only tools. For Claude Code the documented example is: claude mcp add –transport http datagouv https://mcp.data.gouv.fr/mcp. Reconnect and inspect available tools before starting the dataset research pass.
Author
datagouv
Publisher
Community
Last updated
Nov 13, 2026
Quick brief

Use this workflow when an assistant must answer a specific French public-data question with traceable datasets rather than a generic web summary. Input the topic, geography, time period and required output. Connect an MCP client to the official hosted endpoint https://mcp.data.gouv.fr/mcp over Streamable HTTP.

How it works

What this skill actually does

Call search_datasets with the research terms. Inspect promising IDs with get_dataset_info, checking publisher, dates, license and fit to the question. Use list_dataset_resources to enumerate files, then get_resource_info to inspect the selected resource and whether Tabular API access is available. For supported resources, use query_resource_data with an initial small page to inspect column structure and values. Record page limits and distinguish a preview from an exhaustive analysis.

Return a shortlist with dataset and resource IDs, source URLs, publisher and update information, format, license, inspected fields, observed sample rows and remaining gaps. Explain where the data does not cover the requested time or geography. Do not infer complete-population statistics from a sample page. If tabular querying is unavailable, report the resource metadata and link rather than inventing rows.

Invoke this instead of manually navigating the portal when research requires repeated search, metadata comparison and schema inspection inside the current agent conversation. The boundary is a read-only evidence-gathering pass for one question. It excludes publishing datasets, uploading files, bulk ingestion and unrestricted calls to unrelated services. MCP is central to tool invocation, irrespective of the client brand.

The official repository and v1.0.2 README document the endpoint and named tools. Verification covers source metadata and documented invocation; the hosted tools were not executed during intake.