Sirene / SIREN French Company Search — INSEE Register

Stop rebuilding French prospect and due-diligence lists by hand: get companies from the France company register — with their directors (dirigeants), their SIREN and SIRET, and the exact date they were registered — as a spreadsheet, or as a feed that refreshes every morning.

This Actor calls the French government's own open-data API (recherche-entreprises.api.gouv.fr, fed by INSEE/SIRENE and INPI/RNE), so what you export is the register itself: no API key, no login, no broken selectors.

Base identity data on french companies is public everywhere. What most sources do not give you is the named executives, a creation-date filter that surfaces newly registered companies, and the official labels that let you segment a list before contacting anyone.

What you get

Every record is one company, with up to 36 fields:

• SIREN (9 digits) and the head-office SIRET (14 digits), plus the intra-community VAT number
• Executives (dirigeants) — role, name and birth year for individuals, or company name and SIREN for corporate officers. INSEE alone does not publish these.
• Official labels — certified organic, RGE, Qualiopi, training provider, social economy, mission-driven, public service and more
• Creation date and closure date, plus an active flag — the basis for tracking newly registered businesses
• Legal name, trade name, legal form (SAS, SARL, SA, auto-entrepreneur…) with its INSEE code
• NAF/APE activity code and a readable activity label
• Full postal address, postal code, city, department, region, and latitude/longitude of the head office, where published
• The establishment that matched your filters — its SIRET, address, postal code and city, plus up to ten more matching establishments, which is all the source returns per company. A department search matches any establishment, so this is the address inside the department you asked for; the head-office one often is not
• Employee bracket and its reference year, INSEE size category (SME, mid-cap/ETI, large enterprise), sole-trader flag
• Number of establishments, total and currently open
• Filed financials where the company publishes them: revenue and net result, per year
• Last INSEE update timestamp, the source attribution string, and a link to the official annuaire-entreprises.data.gouv.fr page

Who uses this

• A sales team pulls the companies registered in their department and activity in the last 30 days and calls them before anyone else does — using the matched-establishment address, which is the one in their department.
• A compliance officer resolves a SIREN or SIRET to the full legal identity, the officers behind it and the filed accounts, for a KYB file.
• A market analyst maps french business data by activity code and region, with coordinates ready for a GIS or a dashboard.
• A researcher builds a segmented company database of France — all Qualiopi training providers, or all certified-organic producers — from official labels rather than guesswork.
• A developer or AI agent enriches a CRM list of identifiers, or calls the Actor as an MCP tool to answer questions about entreprises France on demand.

Example output

{
  "siren": "478455793",
  "siret": "47845579305205",
  "name": "BOULANGERIES BG (B B G)",
  "legalForm": "SAS - simplified joint-stock company",
  "legalFormCode": "5710",
  "nafCode": "56.10C",
  "nafLabel": "Food and beverage service",
  "vatNumber": "FR94478455793",
  "isActive": true,
  "creationDate": "2004-08-25",
  "closureDate": null,
  "employeeRange": "5000-9999 employees",
  "sizeCategory": "Large enterprise",
  "isSoleTrader": false,
  "address": "615 AVENUE DE LA CHAFFINE 13160 CHATEAURENARD",
  "postalCode": "13160",
  "city": "CHATEAURENARD",
  "department": "13",
  "latitude": 43.89036,
  "longitude": 4.82613,
  "matchedSiret": "47845579301014",
  "matchedAddress": "ZONE COMMERCIALE VAUBAN RUE DE GUESNAIN 59500 DOUAI",
  "matchedPostalCode": "59500",
  "matchedCity": "DOUAI",
  "establishments": 601,
  "openEstablishments": 579,
  "labels": null,
  "executives": [
    { "role": "Président de SAS", "type": "personne physique", "lastName": "…", "firstName": "…", "birthYear": "1956" }
  ],
  "sourceUrl": "https://annuaire-entreprises.data.gouv.fr/entreprise/478455793",
  "source": "recherche-entreprises.api.gouv.fr (INSEE/INPI open data)"
}

Officer names come back in full; they are elided in this sample. labels is null here because this particular company carries none of the official qualifiers; it lists them — certified organic, Qualiopi, RGE, training provider and the rest — for the companies that do. The matched columns are filled when a filter matched a specific establishment; a plain SIREN lookup leaves them null. Labels, legal forms and employee brackets are emitted in English, as shown. matchedEstablishments (omitted here for length) repeats the matching establishments the source hands back, with their own SIRET, address and coordinates. The source returns at most ten per company and gives no way to ask for more: this company has 601 establishments and the export carries ten of them (measured 2026-08-21), which is why the matched address is a shop in Douai (59) while the head office is in Châteaurenard (13). For a single-site company the list is complete; for a chain it is a sample. matchedSiret / matchedAddress describe the first of the matches, and when you searched by department every match returned is inside that department, so the truncation does not affect them. financials carries one entry per published year and is empty for companies that do not file.

Input

Field	What it does
query	Company name, SIREN, SIRET or keywords. Leave empty to browse with the filters below.
sirenList	Exact identifiers to fetch (9-digit SIREN or 14-digit SIRET). This is a lookup mode, not a filter: when it is non-empty the Actor fetches exactly these identifiers and applies no other criterion. query, departments, nafCodes, legalStatus, minEmployees and createdAfter are all ignored — the run log names the ones it dropped — so a company that contradicts them is still returned and still charged. Run the filters as a separate run if you meant to search. Identifiers the register does not resolve are named in the log and counted in the run's status message.
departments	Department codes to filter by, e.g. 75, 13, 2A. Empty = all France. Matches a company when any of its establishments is there — read matchedAddress, not address, for the location in your department.
nafCodes	NAF/APE activity codes, e.g. 47.11F supermarkets, 62.01Z software. Empty = all activities.
legalStatus	Restrict to active (default) or ceased companies. all is not a criterion the source accepts on its own: combine it with another filter.
minEmployees	INSEE employee-bracket floor. Leave it empty for "no minimum": any value, 0 included, keeps only companies whose headcount bracket is published and drops every company whose bracket is NN (not disclosed), which was most of the result set in every search measured below. The filter asks the source for the brackets at or above your floor, and NN is not one of them. Measured 2026-08-21: active software companies in Corsica (2A) are 202 unfiltered and 18 with minEmployees: "0", because 184 of the 202 are NN; restaurants in Lozère (48) go 189 → 35, with 154 NN. Where a value is what you want, it must be a bracket boundary: 0, 1, 3, 6, 10, 20, 50, 100, 200, 250, 500, 1000, 2000, 5000, 10000. Anything larger means the 10,000+ bracket, the largest the register has. A value in between — 15 — is refused with an error instead of being rounded: rounding 15 up to the 20+ bracket drops the 10-19 bracket, and on active Paris software companies that discards 429 of 1,051 (measured 2026-08-21), some of which do have 15 or more employees.
createdAfter	Only companies created on or after this date (YYYY-MM-DD, a real calendar date — 2026-02-30 is refused rather than quietly matching nothing). Filtered by the Actor — see the note below.
scanLimit	How many search results to read through while applying createdAfter. Defaults to 10,000, which is everything a single search can return. Only matching companies are exported and charged.
onlyWithExecutives	Keep only companies whose directors are published.
includeExecutives	Include directors and officers. Default on; turn it off if you only need company-level data.
maxItems	Safety cap on the companies read from the source. 0 = unlimited. Default 200. With onlyWithExecutives on, companies dropped for having no published directors are read and counted against the cap without being exported, so the export can be smaller than the cap: {"query": "carrefour", "onlyWithExecutives": true, "maxItems": 5} exported 3 companies and skipped 2 (measured 2026-08-21). Only exported companies are charged, and the status message says how many were skipped. Raise the cap if you need a set number of companies with executives.

Software companies registered in Bouches-du-Rhône since the start of the month, with a named director to call:

{ "departments": ["13"], "nafCodes": ["62.01Z"], "createdAfter": "2026-08-01", "onlyWithExecutives": true, "maxItems": 0 }

Pair createdAfter with filters narrow enough that the search reports fewer than 10,000 matches, and check the run's status message for PARTIAL COVERAGE: see How the creation-date filter works below for why both matter.

Software companies in Paris with at least 10 employees:

{ "nafCodes": ["62.01Z"], "departments": ["75"], "minEmployees": "10", "legalStatus": "active" }

Enrich identifiers you already hold, company data only:

{ "sirenList": ["478455793", "50393256800010"], "includeExecutives": false }

Run it on a schedule

Save your filters as a Task, set createdAfter to a recent date, and schedule it daily. Each run returns the companies registered since that date, so the dataset becomes a rolling feed of newly registered companies rather than a one-off export. Watch the run's status message: if it ends in PARTIAL COVERAGE the feed missed part of that day's registrations, and the message says how many records went unread — see How the creation-date filter works for the two reasons that happens and what to narrow.

From the run, Apify integrations push the results straight to Google Sheets, Slack, e-mail, a webhook, Make or Zapier — or into your CRM through the webhook. Deduplicate on siren: it is stable, unique per company and never reused, so a daily run appends only genuinely new records.

Use it from code or from an AI agent

Start runs and read the dataset through the Apify API or CLI in any language; results come out as JSON, CSV, Excel or XML. The Actor is also exposed as an MCP tool through the Apify MCP server, so an AI agent can query the French company register directly — by name, by SIREN, or by department and activity.

Pricing

Pay per event: a small fee for each company exported, and nothing for a run that returns no matches. Platform usage on top of that is negligible, because the Actor makes plain API calls and never starts a browser. Current rates are on the Actor's Pricing tab.

How the creation-date filter works

The government's search API has no creation-date parameter — that is a limit of the source, not of this Actor, and it silently ignores any parameter it does not know. So createdAfter is applied here instead: the Actor reads through the search results and exports only the companies created on or after your date. You are charged for exported companies, not for scanned ones.

Three consequences worth knowing before you rely on it:

• Results are ranked by company size, never by date. The register hands out the biggest, oldest companies first, and its only sort parameter — sort_by_size, a boolean — sorts by number of establishments, so there is no way to ask for newest-first. Recent registrations therefore sit deep in the stream. In departments: ["75"] + nafCodes: ["62.01Z"], the first company created since 2024 is the 248th record read: a scan stopping at 200 exports none of them, one stopping at 300 exports 12 (measured 2026-08-21). That is why scanLimit defaults to the full 10,000.
• The source truncates any single search at 10,000 results, and that ceiling cannot be raised. A search reporting exactly 10,000 matches is showing you a slice, not the set — Paris software companies do this, so even a department plus an activity code is not always enough, and a full scan of it finds 2,994 post-2024 companies out of an unknown larger true number (measured 2026-08-21). Narrowing until the reported total is below 10,000 gets you most of the way, but not all: see the next point.
• The source can also stop delivering before the total it advertises. departments: ["13"] + nafCodes: ["62.01Z"] reports 3,972 matches across 159 pages, yet returns nothing from page 158 onwards, so a complete scan reads 3,925 of them and 47 are unreachable by any paging — 1.2% of that search, measured 2026-08-21. The withheld tail is the small-company end, which is exactly where new registrations live: that scan exports 22 companies created since 2026-08-01, and the 47 it never saw could hold more. Splitting the search further (by NAF sub-code, or department by department) is the only way to reach them.

The run log warns whenever a search hits the 10,000 cap, the scan stops early, or the source delivers fewer records than it reported, and the run's status message ends with PARTIAL COVERAGE and the numbers in each case. A run that does not say PARTIAL COVERAGE read every record the source handed over, and the source handed over as many as it said it had.

If you need every new French registration nationwide rather than a filtered slice, the registration announcements in BODACC are the right source for that, and the register search is the wrong tool for it.

Where the data comes from

One practical ceiling to know about: a single search returns at most 10,000 records — the source refuses to page past that, whatever the page size. For a bigger sweep, split the run by department or by activity code: each slice is its own search with its own 10,000-record window.

recherche-entreprises.api.gouv.fr is the official search API operated by the French government (DINUM), built on the SIRENE database (INSEE) and the Registre National des Entreprises (INPI). Both are published as open data and may be reused, including commercially, under the French open licence, with attribution — every record carries the source field for exactly that.

Honest limits. The administration refreshes the data rather than streaming it, so treat it as daily, not real-time. A handful of records carry an implausible creation date (occasionally a future one); those come from the register itself and are passed through unchanged rather than silently dropped. There are no e-mail addresses or phone numbers: the register does not publish them. Coordinates, region, employee year, financials and executives appear only where the source publishes them, and are null otherwise. Companies that opted out of public distribution (statut_diffusion: "P") are not excluded by the source: they come back like any other record, but with their address withheld. address and postalCode then carry the source's literal marker [NON-DIFFUSIBLE] instead of a value — as can matchedAddress and matchedPostalCode, though those prefer a matched establishment whose address is published — while the identity fields (SIREN, name, activity, dates) are normal. In a sample of 200 active Paris companies read on 2026-08-21, 2 were in this state (SIREN 339930984 and 312285901); both still carried their municipality in city. The Actor passes the marker through unchanged rather than blanking it, so you can tell a withheld address from a missing one; drop those rows with postalCode != "[NON-DIFFUSIBLE]" if your pipeline needs real addresses only.

Officer names are published by the French register itself for corporate-transparency purposes; set includeExecutives: false if you do not need them. This Actor is independent and not affiliated with the French administration.

Frequently asked questions

Where does this data come from and can I reuse it?

From recherche-entreprises.api.gouv.fr, the French government's own search API built on INSEE's SIRENE database and INPI's Registre National des Entreprises. Both are open data, reusable commercially under the French open licence with attribution.

Can I get companies created in the last 30 days?

Yes, with createdAfter — but read the note on this page first: the source API has no creation-date parameter, so the France Company Register Scraper filters the results itself. Pair it with a department and an activity code so the scan covers the whole set.

Are directors' names included?

Yes, where the register publishes them, because the French register publishes them for corporate-transparency purposes. Set includeExecutives: false if you do not need them.

Why are there no e-mail addresses or phone numbers?

The register does not publish them. This Actor returns what the source contains and never invents contact data.

Related Actors

Same engine, same official-source approach:

• AusTender (Australia) — federal contracts, winners and expiry dates
• UK Public Tenders — Contracts Finder and Find a Tender in one run
• Swiss Public Tenders (simap.ch) — tenders and awards with prices, bid counts and weighted criteria
• Swiss Commercial Register (SHAB) — new companies, changes, deletions and bankruptcies
• EU Tenders (TED) — European procurement notices and contract awards
• GLEIF LEI — LEI lookup, lapsed-LEI screening, parent companies
• Brazil CNPJ — Receita Federal company data and partners