TrustyData MCP Server
Connect your LLM (Claude, ChatGPT, IDE assistants…) to TrustyData — French address data quality, geocoding, routing, company lookup (SIRENE) and catchment-area statistics (INSEE), built on official open data sources (BAN, INSEE, SIRENE, IGN, OpenStreetMap).
▶️ Try it live — public demo agent, no account needed Ask a real agent to verify an address, list communes, look up a company or compute a route: it calls this MCP server in front of you.
This is a hosted, remote MCP server — there is nothing to install or run. Point your MCP client at the endpoint below and sign in with your TrustyData account.
| Endpoint | https://mcp.trustydata.app/mcp |
| Transport | Streamable HTTP |
| Auth | OAuth 2.1 (sign in with your TrustyData account) |
| Docs | https://trustydata.fr/docs/ |
| Status | https://status.trustydata.app |
This repository hosts the public listing metadata for the TrustyData MCP server (
server.json, documentation, examples). The server itself is a hosted service — the source of the underlying data API is not part of this repo.
What it does
TrustyData turns messy French addresses into clean, authoritative data and adds geographic context, directly inside your LLM conversation:
- Verify & normalize a French postal address against the official BAN reference (the result is authoritative — an empty result means no match).
- Search addresses and localities by name, postal code or INSEE code, with optional filters (department, region, population).
- Proximity search — find addresses or points near a location.
- Routing — compute a road route or a travel-time/distance matrix in France (OpenStreetMap), by car, on foot or by bike.
- Company lookup — search French companies and establishments in the official SIRENE registry and get full records (executives, finances, collective agreements, depending on your plan).
- Catchment areas — population, households, income, age and socio-professional profile, spending potential of a drive-time or radius zone (INSEE Filosofi & census); compare up to 10 zones and measure what each one covers exclusively.
- Activity mix — how many SIRENE establishments a catchment area holds, by family and sub-family of activity, with a density per 10,000 inhabitants and an index against a reference territory: France by default, or the zone's dominant département when you ask for it. A floor, never a total: only active, geolocated establishments are counted, and the response says which reference was used and what share of it that count represents.
- Business failures — collective procedures published in the BODACC (liquidation, receivership, safeguard), counted on every establishment of the company located in the area, and activity stops (SIRENE closures that are neither a takeover at the same address nor a relocation), by year since 2017 and by activity, with a rate per 100 active establishments and an index against the reference territory. Procedures are a floor, stops a ceiling, and the current year is partial — the response says so.
Richer fields (e.g. INSEE Filosofi statistical grid, Lambert 93 coordinates) are returned depending on your plan.
Tools
| Tool | What it does | Minimum plan |
|---|---|---|
verify_address | Verify & normalize a French address against the BAN | Discovery |
search_address | Autocomplete / search full addresses | Discovery |
get_address_details | Full detail for a given address id | Discovery |
search_locality | Search French communes / localities | Discovery |
search_nearby | Proximity search around a point | Growth |
route_matrix | Travel-time / distance matrix (car, foot, bike) | Growth |
compute_route | Full road route between points (car, foot, bike) | Business |
search_company | Search French companies & establishments in the SIRENE registry (name, SIREN/SIRET, activity, location) | Discovery (proximity search: Growth) |
get_company_details | Full record of an establishment or company (identity, executives, finances, collective agreements — by plan) | Discovery |
zone_stats | Everything about a catchment area (drive time or radius): population, households, income, age & socio-professional profile, spending potential, DVF real-estate sale prices (5-year trend, vs. département) and establishment counts by activity (SIRENE, density per 10,000 inhabitants, index vs. France or the dominant département), business failures and activity stops since 2017 (BODACC + SIRENE) | Growth |
zone_compare | Compare 2–10 catchment areas: exclusive population, pairwise overlaps, ranking — each with its full zone_stats block, real-estate prices and activity counts included, business failures and activity stops since 2017 (BODACC + SIRENE) | Growth |
All tools are advertised to every client. If your plan doesn't cover a tool, it returns an actionable upgrade message instead of failing silently.
Prompts
Beyond tools, the server ships five prompts — guided starters that fill in
the right parameters for you. In Claude Code they show up as
/mcp__trustydata__<name>; other MCP clients surface them their own way.
| Prompt | Arguments | What it does |
|---|---|---|
zone_de_chalandise | adresse, minutes (10), mode (car) | Full INSEE profile of a catchment area, with real-estate sale prices and activity counts |
comparer_emplacements | adresses (2–10), minutes (10) | Compare locations, ranked on exclusive population |
potentiel_commerce | adresse, secteur, minutes (10), coefficient | Estimated disposable income and sector spending potential |
verifier_adresses | adresses | Verify a list of addresses, one call each |
qualifier_entreprise | nom_ou_siret | Find a company in SIRENE and open its full record |
Prompt names and wording are French, like the data they describe. A prompt returns a message, not a result: it calls no tool by itself and consumes no quota — the tool it points at applies its own plan requirement when the assistant calls it.
Connect
Claude (claude.ai / Claude Desktop)
Add a custom connector pointing to:
https://mcp.trustydata.app/mcp
You'll be prompted to sign in via OAuth with your TrustyData account the first time a tool is used.
Generic MCP client (Streamable HTTP)
{
"mcpServers": {
"trustydata": {
"url": "https://mcp.trustydata.app/mcp"
}
}
}
Don't have an account yet? Start free on trustydata.fr — the Discovery plan is free (5,000 requests/month, no card required).
Example prompts (free-form)
No prompt needed — just ask:
Vérifie et normalise cette adresse : "1 rue de Rivol 75001 Pari"
Quelle est la population de la commune de Bourg-en-Bresse ?
Trouve les adresses proches du 2 avenue de la Gare à Annecy.
Calcule l'itinéraire routier entre Lyon Part-Dieu et l'aéroport Saint-Exupéry.
Trouve les entreprises de plomberie à Annecy et donne-moi la fiche du premier établissement.
Combien de personnes habitent à 15 minutes en voiture du 1 rue Scribe, 75009 Paris, et quel est leur niveau de vie ?
Compare 10 minutes en voiture autour de la Part-Dieu et autour de Bellecour : quelle zone capte le plus d'habitants qu'elle est seule à couvrir ?
Combien d'établissements de restauration y a-t-il à 10 minutes à pied de la place Bellecour, et est-ce beaucoup pour le nombre d'habitants ?
Data sources & attribution
TrustyData relies on official, regularly-updated open data. Please keep the attributions returned by the tools:
- Addresses & communes — Base Adresse Nationale (BAN) & INSEE
- Companies — SIRENE (INSEE), Registre National des Entreprises (INPI)
- Statistical context & catchment areas — INSEE Filosofi and census, IGN Contours IRIS
- Real-estate sale prices — DGFiP, Demandes de valeurs foncières (Etalab, open licence)
- Establishments by activity — SIRENE (INSEE), active geolocated establishments only
- Business failures — BODACC (DILA) and SIRENE (INSEE), geolocated establishments only
- Routing — OpenStreetMap contributors (ODbL)
Documentation
Full API guides, endpoint reference and examples: https://trustydata.fr/docs/
License
The contents of this listing repository (documentation, server.json, examples)
are released under the MIT License. The hosted service and the
underlying data API are operated by TrustyData and governed by its terms.