For AI agents
Official company records your agent can use, and cite.
Fuentio runs an MCP server, so Claude, Cursor and any MCP client can verify a company against its official register in the middle of a task. Every answer starts with one plain sentence and ends with its source.
Built for AI agents: one MCP server and one REST API, with honest gaps: what isn't covered is said, never guessed.
Connect it to your agent
-
Claude Code
One command, no key needed to start.
claude mcp add --transport http fuentio https://api.fuentio.com/mcpWith a key
claude mcp add --transport http fuentio https://api.fuentio.com/mcp \ --header "Authorization: Bearer YOUR_API_KEY" -
Claude Desktop and claude.ai
- Settings → Connectors → Add custom connector.
- Name: Fuentio. URL: the address below. Then Add.
- In a chat, turn on Fuentio from the tools menu and ask, for example, “Is SIREN 441639465 still active?”
Custom connectors can't send a key, so they use the keyless trial.
https://api.fuentio.com/mcp -
Cursor
In ~/.cursor/mcp.json, or .cursor/mcp.json in a project.
{ "mcpServers": { "fuentio": { "url": "https://api.fuentio.com/mcp" } } }With a key
{ "mcpServers": { "fuentio": { "url": "https://api.fuentio.com/mcp", "headers": { "Authorization": "Bearer YOUR_API_KEY" } } } } -
VS Code (GitHub Copilot agent mode)
In .vscode/mcp.json. With a key, VS Code asks for it once and stores it securely.
{ "servers": { "fuentio": { "type": "http", "url": "https://api.fuentio.com/mcp" } } }With a key
{ "inputs": [ { "type": "promptString", "id": "fuentio-key", "description": "Fuentio API key", "password": true } ], "servers": { "fuentio": { "type": "http", "url": "https://api.fuentio.com/mcp", "headers": { "Authorization": "Bearer ${input:fuentio-key}" } } } } -
ChatGPT (developer mode)
- Settings → Apps & Connectors → Advanced settings: turn on Developer mode.
- Create a connector: name Fuentio, MCP server URL: the address below, authentication: No authentication.
- In a chat, pick the connector from the tools menu and ask your question.
Menu names change often; ChatGPT's help pages have the current path. ChatGPT connectors use the keyless trial.
https://api.fuentio.com/mcp -
Any streamable-HTTP client
Stateless, JSON responses. No key needed to start.
POST https://api.fuentio.com/mcp Accept: application/json, text/event-streamWith a key
POST https://api.fuentio.com/mcp Accept: application/json, text/event-stream Authorization: Bearer YOUR_API_KEY
The tools Live
-
list_coverage()FreeWhich countries and identifier formats are covered, and how fresh the data is. Call it first when unsure.
-
lookup_company(country, identifier, id_scheme?, format?)1 unitVerify a company before onboarding, paying or contacting it. France: SIREN, SIRET or VAT number. Spain: Registro Mercantil sheet.
-
search_companies(country, query, postal_code?, locality?, status?, page?, format?)2 unitsFind a company by name when you don't know its number.
-
get_company_history(country, identifier, page?)1 unitSee what changed at a company (status, address, name, legal form) and when, with the official source of each change. Spain also gets its BORME acts as dated events.
-
verify_company(country, identifier, id_scheme?, reference?)5 unitsA KYB check for onboarding, or before paying or signing: an outcome (verified, partial or unverified), the reasons and a proof code. Use lookup_company for a quick read, and search_companies first when you only have a name.
-
resolve_company(country, name?, identifier?, postal_code?, locality?)2 unitsIs this the same company? Match a supplier or customer record (an official ID, or a name with a postcode or town) to the official registers: candidates, each with a decision and the reasons. Only an official ID links; a name is never proof.
-
explain_match(left, right)1 unitWhy two register records are, or are not, the same company: the decision, the reason codes and the evidence.
-
watch_company(entity_id, callback_url)1 unitWatch a company: when its official records change, Fuentio sends the changes, signed, to your https endpoint. Needs an API key.
Export formats
lookup_company and search_companies, and the REST lookup and search, take an optional format. Same units, except a search file with more than 25 rows: 2 units per started 25 rows.
JSONCSVExcelMarkdownHTMLXMLNDJSONParquetTSV
- At most 100 rows per answer.
- Excel and Parquet come back as a download link that works for 24 hours.
- Every row keeps its source and check time.
A search as CSV
curl "https://api.fuentio.com/v1/companies/search?country=FR&q=renault&format=csv"
Which tool when
| Your question | Tool |
|---|---|
| You only have a company's name. | search_companies |
| You have a SIREN, SIRET or VAT number, or a Spanish registry sheet. | lookup_company |
| You are onboarding a company and need a KYB check. | verify_company |
| You want to know what changed, and when. | get_company_history |
| You have a supplier or customer record and need to know which registered company it is. | resolve_company |
| You want to hear when a company changes. | watch_company |
| You are not sure what is covered. | list_coverage |
What your agent sees
A real record from our snapshot, as the lookup tool returns it:
The tool call
lookup_company({"country": "FR", "identifier": "441639465"})
The first line your agent reads
RENAULT (FR, SIREN 441639465) is active as of 2025-12-06, source: Annuaire des Entreprises (API Recherche d'entreprises).
Then the full record as JSON, with its source, licence and check time.
Prefer plain REST?
The same data over HTTPS. Try it without a key, then create a free key in one call.
A free trial without a key (20 units a day), and free keys with 100 units a month.
Look up a company, no key needed
curl https://api.fuentio.com/v1/companies/FR/441639465
Create a free key
curl -X POST https://api.fuentio.com/v1/keys
Errors an agent can act on
invalid_identifier- The number has the wrong format or check digits.
not_found- A well-formed number with no public record.
id_scheme_not_supported- A known kind of ID we don't look up, such as a Spanish NIF: search by name.
quota_exceeded- Units used up; the message says how to continue.
upstream_unavailable- The official source isn't answering; try again in a minute.