SIRENE Scraper: French Companies by NAF, SIREN & SIRET avatar

SIRENE Scraper: French Companies by NAF, SIREN & SIRET

Pricing

from $0.75 / 1,000 company records

Go to Apify Store
SIRENE Scraper: French Companies by NAF, SIREN & SIRET

SIRENE Scraper: French Companies by NAF, SIREN & SIRET

Every French company of a SIRENE segment, past the 10,000-result limit of the official search. Filter by NAF code, department, postal code, size, legal form, revenue and labels, or look up SIREN and SIRET numbers. Officers, accounts, new-company monitoring and webhook.

Pricing

from $0.75 / 1,000 company records

Rating

0.0

(0)

Developer

Alexis Morain

Alexis Morain

Maintained by Community

Actor stats

0

Bookmarked

3

Total users

2

Monthly active users

a day ago

Last modified

Features

Limited permissions

Share

SIRENE Scraper: French companies by NAF code, SIREN and SIRET

SIRENE Scraper returns French companies from the SIRENE business register as clean rows: SIREN, SIRET of the head office, name, NAF activity with its label, legal form, headcount range, category, address and coordinates, directors, revenue and net income, VAT number, labels, and a link to the public record. Pick a NAF code, a zone, a size or a list of SIREN numbers, and it returns every matching company, not the first 10,000. No account, no cookie, no API key.

It reads the open recherche-entreprises API of the French administration (recherche-entreprises.api.gouv.fr), built on SIRENE and other public registers. That API stops at 10,000 results per search and does not say how many companies it left out. This Actor splits a large search into slices small enough to read in full, removes the duplicates the slices create, and returns each company once.

It is built for sales and outbound teams building a target list by activity and region, lead generation agencies, data teams that refresh a company table, and researchers sizing a sector: a SIRENE API alternative that needs no key.

What does SIRENE Scraper do?

The Actor sends searches to the recherche-entreprises API, reads every page of every slice, and writes one row per company to the dataset. You can search in four ways, alone or combined:

  1. By segment: NAF codes (for example 10.71C, bakeries), NAF sections, a text query, a zone (departments, postal codes, commune codes, regions), employee ranges, company categories (PME, ETI, GE), legal forms, revenue, status, and labels such as RGE, Qualiopi, organic, ESS, SIAE, training organization or association.
  2. By number: a list of SIREN (9 digits) or SIRET (14 digits). Each one returns its company, or is reported as not found.
  3. By radius: a latitude, a longitude and a radius up to 50 km, optionally with a NAF code or section.
  4. By time: companies created on or after a date, and monitoring that returns only the companies you have not received before.

What it does not do: it does not return establishments as separate rows (only as a list on each company), it does not give the full SIRENE file, and it does not give emails, phone numbers or websites, which the source does not carry.

Why use SIRENE Scraper?

The full segment, not the first 10,000. A plain search for NAF 43.21A (electrical installation) in the department 93 stops at 10,000. The Actor returned 10,489 unique companies for the same search, in 107 seconds. The section "The 10,000 limit, with numbers" below shows how that works and where the counts differ.

No key. The source is open. The Actor keeps to about 5 requests per second, under the limit of 7 per IP address, and retries refusals with a pause.

Filters you can check. Each filter was tested against the source one at a time: the total moves in the right direction when you add it. A wrong code (a NAF code that does not exist, a department 999) stops the run at once with a clear message instead of returning nothing.

Monitoring and a fair default. Run the same search every week and get only the new companies. Companies whose owner opposed commercial prospecting are skipped and not charged, and you can turn that off (see Defaults you should know).

Scheduling, the API, a webhook at the end of the run and integrations with Make, n8n, Zapier, Google Sheets and Slack come with the Apify platform.

It resumes after a restart: companies already delivered are not delivered or charged twice.

What data can you extract?

Every row has these fields. Missing values are null, never an error. Numbers are numbers, dates are YYYY-MM-DD.

FieldTypeWhat it is
sirenstring9-digit company number
siretstring14-digit number of the head office
namestringName as shown by the registry
legalNamestringRegistered legal name
acronymstringAcronym (sigle), when there is one
nafCodestringMain activity code, NAF revision 2 (for example 43.21A)
nafLabelstringLabel of the NAF code, null if the code is unknown
naf25CodestringMain activity in the NAF 2025 nomenclature
sectionstringNAF section letter, A to U
legalFormCodestringINSEE legal form code (nature juridique), for example 5710
legalFormLabelstringLabel of the legal form, null if unknown
companyCategorystringPME, ETI or GE
employeeRangeCodestringINSEE headcount range code (NN, 00, 01, 02, 03, 11 to 53)
employeeRangeLabelstringThe range in words, for example "20 à 49 salariés"
employeeRangeYearnumberYear the headcount range refers to
isEmployerbooleanTrue when the company has employees
createdAtstringCreation date of the company
statusstringactive or closed
closedAtstringClosing date, when closed
addressstringHead office address
postalCode, city, communeCodestringHead office postal code, city and INSEE commune code
department, regionstringDepartment and region codes of the head office
latitude, longitudenumberHead office coordinates
establishmentsCountnumberNumber of establishments, closed ones included
openEstablishmentsCountnumberNumber of open establishments
directorsarrayOfficers: a person (lastName, firstNames, birthYear, nationality) or a company (name, siren), with their role
revenue, netIncome, financialYearnumberLatest published accounts, in euros
financesarrayRevenue and net income for every published year, latest first
vatNumberstringIntra-community VAT number
collectiveAgreementsarrayIDCC codes of the collective agreements
labelsobjectTrue or false for rge, qualiopi, organic, ess, siae, trainingOrganization, association, soleTrader, livingHeritage, missionCompany, performingArtsEntrepreneur, healthSocial, school, responsiblePurchasing, foodHygieneChecked, lawyer
commercialOptOutbooleanTrue when the company opposed commercial use of its data
updatedAtstringLast update of the record at the source
annuaireUrlstringThe company page on annuaire-entreprises.data.gouv.fr
establishmentsarrayOnly with the option: the establishments that match your filters (SIRET, address, NAF, status, headcount, coordinates), up to 100 per company. Empty when you filter on nothing about establishments: the source returns the ones that matched
scrapedAtstringWhen the row was read

Revenue and net income exist only for companies that publish accounts. In a sample of 25 bakeries, 13 had them. Small companies and sole traders often do not, and the field is null for them.

Results come in the dataset, which you can export as JSON, CSV, Excel, XML or HTML, or read through the API.

How to get a list of French companies by NAF code

  1. Open the Actor and type a NAF code in NAF / APE codes, for example 43.21A. The code can be written 4321A too.
  2. Add a zone if you want one: departments, postal codes, commune codes or regions. Leave it empty for the whole country.
  3. Set Max companies. Start with 50 to look at the rows. The default is 1,000.
  4. Press Start. A segment of a few thousand companies takes under a minute. The test run of 10,489 companies took 107 seconds.
  5. Download the dataset from the Output tab, or read it from the API.

To look up one company, put its SIREN or SIRET in SIREN or SIRET numbers instead.

The 10,000 limit, with numbers

The recherche-entreprises API refuses any page beyond the 10,000th result, and when a search matches more than that, it reports 10,000 as the total. So you cannot tell, from the answer, how many companies you missed. For NAF code 43.21A, a plain search returns 10,000. Adding up the totals the API gives department by department (metropolitan and overseas) gives 234,575 for the same code. A plain search reads about 4 percent of the segment.

The Actor handles it by splitting. When a slice reports 10,000 or more, it is cut again: first by department (if you did not give one), then by employee range (16 codes), then by legal form (260 codes). Only then does it cut by commune, as a last resort. A slice that is still full after that is read to its limit and reported in the run status and the SUMMARY record. Those values come from the API itself, and they add up: for 43.21A in the department 35, the totals by employee range and the totals by legal form both sum to 2,217, the same as the total for the department.

I tried other cuts first and dropped them. Cutting by postal code loses companies whose establishment address is hidden ("[NON-DIFFUSIBLE]"): 24 of 2,217 in the department 35. Cutting by commune with the official commune list loses whole communes the list does not know: in the department 93 it missed one commune and about 180 companies. The employee range and legal form cuts lost nothing in the same tests.

The test that gives the headline number, run on the Apify platform:

SearchResult
Plain search, NAF 43.21A, department 9310,000 (the cap)
This Actor, same search, all statuses10,489 unique companies
Time and requests107 seconds, 435 requests, 86 MB of memory
Sum of the slice totals10,598

Why 10,598 and 10,489 do not match, and why the national figure of 234,575 is not exact either:

  1. Department, postal code and commune filters match any establishment of a company, not only its head office. A company with a shop in Paris and one in Hauts-de-Seine is found in both departments. For bakeries (10.71C), 243 of the 3,753 companies of the department 75 also appear in the department 92. Adding department totals counts those companies twice, so 234,575 is an upper bound for the number of electrical installation companies, and I have not measured the exact national count. The Actor removes the duplicates before it returns anything and before it charges.
  2. The total the API announces counts a little more than it serves. For bakeries in Paris it announced 3,759 and served 3,753. For one slice of the department 93 it announced 8,945 and served 8,836, and a second cut of the same slice by legal form served the same 8,836 companies. So about 1 percent of the announced number is not a missing company, it is a counting difference at the source.

The Actor returns every company the API will serve for your filters, once. The SUMMARY record puts the slice total, the unique companies read and the number returned side by side so you can check a run.

Defaults you should know

Two defaults change how many rows you get. Both are on purpose.

Companies opposed to commercial use are skipped

The registry carries a flag, opposition à la diffusion commerciale, for companies whose owner asked that their data is not used for commercial prospecting. It concerns mostly sole traders. The Actor skips those companies by default (excludeCommercialOptOut is true) and does not charge for them.

How many that is depends on the segment. In my runs it was 21 percent of the active bakeries of Paris (490 of 2,332) and about 4 in 10 for the active bakeries of Ille-et-Vilaine (36 skipped for 50 returned in a test). Expect between a fifth and two fifths for activities with many sole traders, much less for companies with employees.

To get them, set Skip companies opposed to commercial use to off. Then you carry the responsibility: the flag is returned in commercialOptOut for each row, and prospecting a company that opposed it is your decision and your legal risk. If you use the data for something other than prospecting (market sizing, research, compliance checks), turning it off is normal.

Status is active

By default only active companies are returned (status active). The source returns active and closed companies together when you give no status, so the same search gives roughly twice as many rows without this default: for 43.21A in the department 35 the source reports 2,217 companies, of which 1,194 are active and 1,023 closed. Set status to closed for the closed ones, or all for both. For bakeries in the department 75, status all returns 3,753 companies and the default returns the 2,332 active ones.

Monitoring new French companies

There are two ways to follow new companies. They can be combined.

Created after a date. Set Created on or after to a date like 2026-11-01. The source cannot filter on the creation date, so the Actor reads the segment and keeps the companies created from that date on. The companies it skips are not charged. This gives you the companies registered in a period, for a NAF code and a zone.

New only. Switch on New companies only. The first run returns everything that matches and remembers the SIREN numbers. The next runs with the same filters return only companies that were not returned before. Companies not returned for another reason, for example because of a spending limit, come back in the next run. The memory is a key-value store in your account, named sirene-company-scraper-registry by default. Give a Monitoring list name to keep several lists apart.

A typical weekly job: NAF 62.01Z, department 69, created after the first of the month, new only on, a webhook to Slack or n8n.

The monitoring reads the whole segment each time, because the source cannot sort by date. A large segment costs run time each week, and you pay only for the new companies.

Input

See the input tab for every option. The fields:

FieldTypeWhat it does
nafCodeslistNAF rev. 2 codes, OR between values. 10.71C or 1071C
sectionslistNAF section letters A to U
querytextFree text: name, address or officer name
sirenslistSIREN (9 digits) or SIRET (14 digits), one lookup each
departmentslistCodes like 35, 2A, 971. Matches any establishment
postalCodeslist5-digit codes. Matches any establishment
communeCodeslist5-character INSEE codes (75101, 2A004)
regionslistINSEE region codes (11, 53, 93)
latitude, longitude, radiusKmnumbersRadius search up to 50 km. Only NAF code and section can be combined
employeeRangeslistINSEE headcount codes: NN, 00, 01, 02, 03, 11, 12, 21, 22, 31, 32, 41, 42, 51, 52, 53
companyCategorieslistPME, ETI, GE
legalFormslistINSEE legal form codes: 5710 (SAS), 5499 (SARL), 1000 (sole trader), 9220 (association)
excludeSoleTradersbooleanSkip individual entrepreneurs
statustextactive (default), closed or all
minRevenue, maxRevenuenumbersEuros, on the latest published revenue. Companies without accounts are excluded
isRge, isQualiopi, isBio, isEss, isSiae, isTrainingOrganization, isAssociationbooleansLabel filters
createdAfterdateKeep companies created on or after this date. Skipped ones are not charged
excludeCommercialOptOutbooleanOn by default
includeEstablishmentsbooleanAdd the establishments that match your filters
maxItemsnumberStops after this many companies. Default 1,000
newOnlybooleanReturn only companies not returned by earlier runs with the same filters
registryNametextName of the monitoring list
webhookUrlURLReceives a JSON summary at the end of the run

Filters of different fields are combined with AND, values inside one field with OR. A first test:

{
"nafCodes": ["10.71C"],
"departments": ["35"],
"maxItems": 50
}

A radius search (latitude, longitude, radiusKm) is limited to 10,000 companies by the source and cannot be split. Only NAF codes and sections can go with it, and the Actor applies the status after the read.

Output example

One row, with a few fields left out and the officers cut to two:

{
"siren": "895362770",
"siret": "89536277000011",
"name": "YOU&MIE",
"legalName": "YOU&MIE",
"nafCode": "10.71C",
"nafLabel": "Boulangerie et boulangerie-pâtisserie",
"section": "C",
"legalFormCode": "5710",
"legalFormLabel": "SAS, société par actions simplifiée",
"companyCategory": "PME",
"employeeRangeCode": "12",
"employeeRangeLabel": "20 à 49 salariés",
"employeeRangeYear": 2024,
"isEmployer": true,
"createdAt": "2021-03-15",
"status": "active",
"address": "IMMEUBLE COLIVIA - ILOT B8 1 PLACE VIASILVA 35510 CESSON-SEVIGNE",
"postalCode": "35510",
"city": "CESSON-SEVIGNE",
"department": "35",
"region": "53",
"latitude": 48.1325661490261,
"longitude": -1.62037463153767,
"establishmentsCount": 3,
"directors": [
{ "type": "company", "name": "MONTGISON", "siren": "937993780", "role": "Président de SAS" },
{ "type": "company", "name": "SAINT-HONORÉ", "siren": "938016912", "role": "Directeur Général" }
],
"revenue": 0,
"netIncome": 102952,
"financialYear": 2023,
"finances": [{ "year": 2023, "revenue": 0, "netIncome": 102952 }],
"vatNumber": "FR18895362770",
"collectiveAgreements": ["0843"],
"labels": { "organic": true, "rge": false, "qualiopi": false, "ess": false, "association": false },
"commercialOptOut": false,
"updatedAt": "2026-11-17T09:23:46",
"annuaireUrl": "https://annuaire-entreprises.data.gouv.fr/entreprise/895362770",
"scrapedAt": "2026-11-17T10:00:00+00:00"
}

Officers also carry the null fields (lastName, firstNames, birthYear, nationality), and labels lists all sixteen flags. Revenue is what the company filed: this one filed a revenue of 0.

A run also writes a SUMMARY record to the key-value store: for each search, the status, the total announced by the slices, the rows read, the duplicates removed, the companies skipped by each rule, the requests, the splits by level, and any slice that stayed at the cap or could not be read.

How much does it cost to get French company data?

The Actor uses pay-per-event pricing. You pay for the start of the run and for each unique company returned. Prices fall with your Apify plan.

EventFree planStarterScaleBusiness
Actor start (per GB of memory, 256 MB by default)0.002 USD0.002 USD0.002 USD0.002 USD
Company1.50 USD per 1,0001.50 USD1.20 USD0.75 USD

A company found by several slices of the same run is charged once. Companies skipped by your filters, by createdAfter, by the commercial opt-out default or by the monitoring are not charged. A SIREN that does not exist returns nothing and costs nothing. You can set a maximum charge per run in the run options: the Actor stops when it is reached, and the status says so. I checked this: a run with a limit of 0.20 USD returned 132 companies, charged 132 company events and 1 start event, which is 0.20 USD.

Examples at the Free plan price: 1,000 companies cost about 1.50 USD. The 10,489 companies of the 43.21A test in the department 93 cost about 15.73 USD at the Free price and 7.87 USD at the Business price. A weekly watch that returns 40 new companies costs the start fee and 0.06 USD.

The platform usage of a run is small and goes on top only on plans that bill usage separately. The 10,489 company run used about 0.06 USD of platform usage, which is 0.006 USD per 1,000 companies.

Tips and limits

Start small: run 50 companies, look at the rows, then raise Max companies. A whole national segment is large (234,000 companies at the Free price is a few hundred dollars), so set a limit or a maximum charge.

Give a department or a region when you can. The Actor reads about 100 companies per second in the tests, and a zone keeps it off the slices you do not need.

Companies that are not diffusible are not in the source: the registry API leaves out companies whose owner asked for full non-disclosure. Some companies are partly hidden, and the establishment fields can read [NON-DIFFUSIBLE]. Those companies are returned.

The data is as fresh as the register. The updatedAt field shows when the source last changed the record. I did not measure how long a new company takes to appear.

Use it from code

Python:

from apify_client import ApifyClient
client = ApifyClient("<YOUR_APIFY_TOKEN>")
run = client.actor("alexmorain/sirene-company-scraper").call(run_input={
"nafCodes": ["43.21A"],
"departments": ["93"],
"employeeRanges": ["11", "12", "21"],
"maxItems": 5000,
})
for company in client.dataset(run["defaultDatasetId"]).iterate_items():
print(company["siren"], company["name"], company["city"], company["employeeRangeLabel"])

JavaScript:

import { ApifyClient } from 'apify-client';
const client = new ApifyClient({ token: '<YOUR_APIFY_TOKEN>' });
const run = await client.actor('alexmorain/sirene-company-scraper').call({
nafCodes: ['62.01Z'],
departments: ['69'],
createdAfter: '2026-10-01',
newOnly: true,
registryName: 'dev-studios-lyon',
});
const { items } = await client.dataset(run.defaultDatasetId).listItems();
console.log(items.length, items[0]?.name);

cURL, synchronous call that returns the rows:

curl -X POST "https://api.apify.com/v2/acts/alexmorain~sirene-company-scraper/run-sync-get-dataset-items?token=<YOUR_APIFY_TOKEN>" \
-H "Content-Type: application/json" \
-d '{"sirens": ["552032534", "79491190900024"], "excludeCommercialOptOut": false}'

Frequently asked questions

Is there a SIRENE API without a key?

Yes. The recherche-entreprises API used here needs no key and no account. It limits you to 7 requests per second per IP address, 10,000 results per search and 25 results per page. This Actor handles the rate and works around the 10,000 limit. The official INSEE SIRENE API needs a key and gives other fields (the full file); this one is simpler and carries officers and published accounts.

How do I export all companies by NAF code?

Put the NAF code in NAF / APE codes, keep Max companies high enough, and start. A company found in several departments is returned once. A list of SIREN or SIRET numbers works too, one lookup per line. For a code with more than 10,000 companies, the Actor splits the search by department, employee range and legal form and returns them all. Export the dataset as CSV or Excel. To keep the run short for a big code, add a region or departments.

How can I get a list of new companies in France?

Use Created on or after with a date, with or without a NAF code or zone, and switch on New companies only for the following runs. Schedule the run weekly and add a webhook. The first run returns the companies created since your date, the next ones return only the companies that appeared since.

Why did I get fewer companies than the source announces?

The usual causes are a limit you set (Max companies or a maximum charge), the commercial opt-out default, the status default, createdAfter, or the new-only memory. The run status lists what was skipped by each rule. After that, the announced number is about 1 percent above what the source serves.

Does it give emails, phone numbers or websites?

No. SIRENE does not carry them. The row gives the legal and location data, the officers and the published accounts, and the annuaireUrl link to the public page. To find a contact, combine the SIREN list with a tool that enriches by company name or domain.

Does it return the officers' personal data?

Officers are in the public register, with last name, first names, birth year and role. The Actor returns what the source returns. Personal data is protected by the GDPR, so keep that data only when you have a legitimate reason. The commercial opt-out default exists for that reason.

Can I use the data commercially?

The data comes from the SIRENE base and public registers under the Licence Ouverte: reuse is allowed with attribution of the source (INSEE and the French administration). You remain responsible for how you use it, including prospecting rules and the opposition flag. This Actor is not affiliated with INSEE or the French administration.

Something is wrong or missing

Open an issue on the Actor page. A sample input and the run link help, and fixes are fast.

Our Actors only extract what is public. However, your results could contain personal data (the names of officers). Personal data is protected by the GDPR in the European Union and by other regulations around the world. Do not collect personal data without a legitimate reason. If you are unsure, consult your lawyers.

Source: SIRENE base (INSEE) through recherche-entreprises.api.gouv.fr, Licence Ouverte. The API tab shows how to start runs from your code.

Guide and use cases

Use cases for outbound and automation, the output fields, pricing and common questions are on the SIRENE Scraper page of my site. The Lead generation family page puts it side by side with the other scrapers of the same family.