Find prospects that look like your best customers
“Find 50 companies that look like these five customers, in Auvergne-Rhône-Alpes.” You know your best customers, rarely their activity code or headcount band. Give Claude 2 to 5 of them, by their SIREN, and ask: Sirenic reads their records in the official registers, deduces a common profile (activity, headcount, legal form, creation date, area) and returns the active companies that match it, your customers excluded. The profile comes back as is, so you can read it, narrow it or widen it yourself.
See a response on test data (free) Get the quote for /v1/prospection/sosies
What the response says
Claude calls the find_french_lookalike_companies tool of Sirenic's public MCP server (https://api.sirenic.eu/mcp), which serves the /v1/prospection/sosies route. The response gives:
- the deduced profile, in the form of the criteria of the
/v1/entreprises/requete/listerroute (profil.criteres): the NAF codes of the customers kept, the range of their headcount bands, their legal forms, their creation dates widened by 2 years on each side, and the area; - the deviations (
profil.ecarts), from a closed list, each with the rank of the customers it concerns in your list, never their SIREN:exemple_introuvable,exemple_non_personne_morale,exemple_cesse,naf_hors_rev2,naf_heterogene,zone_heterogene,effectif_non_diffuse,creation_absente,zone_absente; - the widening steps (
profil.elargissements): under 50 results, the profile widens one step at a time, in this order:naf_division, the common NAF division instead of the exact codes;tranches_un_cran, the headcount bands, one more step on each side;familles_de_formes, the common family of legal forms;zone_region, the region, when the deduced area is a departement. Each step applied is stated with the total before it, a step that does not apply is skipped, and the area you give is never widened; - the first page of 100 companies and the exact total; the next pages are read through
/v1/entreprises/requete/lister, with the same criteria.
Every company served matches the whole profile; none is ranked by likeness. Your customers come back neither on the first page nor on the next ones.
A measured example
Read on 2026-10-17 on the production service, with our test key. Five example customers: five active companies in Auvergne-Rhône-Alpes (region=84), with the same legal form (5710) and the same NAF code (62.01Z), in headcount bands 11 and 12 (10 to 49 employees), created from 2013 to 2018. Answer in 1.3 s: the five customers kept, no deviation, no widening step; 85 companies match the profile, all on the first page, and none of the five customers is among them.
The same day, five other customers in the same region, from five different NAF divisions, in headcount bands 21 and 22 (50 to 199 employees): answer in 1.6 s. The naf_heterogene deviation is stated for all five; the steps applied are stated one by one, tranches_un_cran (total before: 0), then familles_de_formes (total before: 2); naf_division and zone_region do not apply (no common division, and the region given is never widened). Two companies match the widened profile, both in the region asked for.
The SIRENs of the example customers are not published here: the shape of the call is on the route's reference page. Each of these calls is one call of a verified account's free quota; outside it, it costs $0.02.
Limits
- 2 to 5 customers per call, distinct SIRENs with a valid check digit; otherwise an
exemples_invalidesrefusal, which states the ranks at fault if any, never billed. - One area at most,
region(INSEE region code) ordepartement, never both; it is never widened. Without an area, the profile takes your customers' common departement, otherwise their common region, otherwise their regions. - A customer that is missing, is not a legal entity whose register entry is public, or has ceased is set aside and stated by its rank; if none remains, a
profil_impossiblerefusal, never billed. - 15 s for the whole call, reading the customers and the widening steps included; beyond that, a
requete_trop_largerefusal, never billed.
What the route does not do
It adds no contact: no e-mail address, no phone number, no named contact; that data is not in the official registers. It does not rank by likeness and does not read activity signals (hiring, growth, contracts won): the profile rests on the registers alone. No sole trader enters the profile or the results: active legal entities only.
Price
$0.02 per call: the profile, the first page of 100 companies and the exact total. This route is part of an account's 150 free calls a month; beyond them, it is paid with prepaid credits (card top-up is not open yet), or over x402. The next pages are read through /v1/entreprises/requete/lister, $0.02 per page. A refused or unavailable call is never billed.
How you pay
Every route is paid per call, in USDC or EURC on the Base network, over the x402 protocol: no account to create, no API key, no subscription. The first call returns a 402 quote your client settles, then replays the call. A failed call is never billed.
Open today: 150 free calls a month with an account, on 109 of 125 routes (the price list says which), and pay-per-call over x402 (USDC or EURC), with no account except registry filings. Prepaid credits: card top-up is not open yet.
Paid responses are Ed25519-signed: you can later prove what was served to you, and when. The whole catalogue is free to read in the OpenAPI document and in llms.txt.
One call, end to end
# x402 client (npm): the 402 quote is settled and the call replayed automatically
npx x402-fetch https://api.sirenic.eu/v1/entreprise/552032534
# Or over MCP, in Claude Code / Cursor
claude mcp add --transport http sirenic https://api.sirenic.eu/mcp
# Or as a local server (stdio), with your key or an x402 wallet
claude mcp add sirenic -e SIRENIC_API_KEY=srn_live_... -- npx -y sirenic-mcp
Routes used and pricing
| Route | Price | Sample | What it returns |
|---|---|---|---|
/v1/prospection/sosies | $0.02 | free sample | Find French companies that look like your best customers: give 2 to 5 customer SIRENs (?exemples=), optionally a region or departement, and get the first page of 100 active legal persons matching the profile deduced from them (NAF codes, workforce brackets, legal forms, creation dates, area), customers excluded, with the exact total. Under 50 results the profile widens one step at a time, each step stated; it is served as criteria to read the next pages. |
/v1/entreprises/requete/lister | $0.02 | free sample | List French companies matching free cross-criteria on a pre-joined table of the official registries (the same criteria as the count route): geography, NAF activity, legal form, workforce, age, last filed accounts, public signals. Active legal persons by default. 100 companies per page, sorted on identite.siren, finances.date_cloture, finances.ca, finances.resultat_net, finances.bfr_sur_ca, with the exact total and the cut stated; each page is one payment. |
/v1/prospection | $0.02 | free sample | French B2B prospecting and lead generation over the full official French company registry: build targeted company lists with multi-criteria filters: NAF/APE activity code, departement or postal-code prefix, legal form, workforce bracket, company age, active RGE environmental certification, gender-equality index. Returns 100 active companies per page; each page is one payment. Combine filters to narrow a market segment into fresh, registry-grade company lists for France. |
Frequently asked questions
I only have my customers' names: what do I do?
Find their SIREN first: the page “Enrich and clean your CRM” returns it for every record of your export, from the name and the head-office postcode. Then give the tool the SIRENs of 2 to 5 of your best customers.
Are the results ranked by likeness?
No. Every company matches the whole profile, and the list follows the order of the /v1/entreprises/requete/lister route. Keep the profile served: add a criterion to narrow the list, remove one to widen it, then read it again through /v1/entreprises/requete/lister.
And beyond the first page?
The response states the exact total. The next pages are read through /v1/entreprises/requete/lister with the criteria of profil.criteres, $0.02 per page of 100; your customers stay excluded.
Take it further
No account to create, no key to request: the first step is free and shows you exactly what you would be buying.
See a response on test data (free) Get the quote for /v1/prospection/sosies Read the route reference Create an account: 150 free calls a month