The question is not “what is their score?” but “why this score, and on which accounts?”. A score computed from 2017 accounts is worthless in 2026, and a score that does not return its scorecard cannot be challenged. /v1/score/defaillance/{siren} computes at call time and returns its entire reasoning.
See a real response (free) Get the quote for /v1/score/defaillance/{siren}
The default-risk score is a decision-support indicator. It is neither a solvency opinion nor a credit rating in the regulatory sense, and it guarantees no payment. It returns its scorecard, every component with its threshold, and a confidence level — so that you can challenge it line by line rather than believe it.
This is the costliest trap in reading filed accounts, and it was fixed here after measurement. When the upstream source publishes a ratio as 0 without having computed it — because the line is missing from the tax form, not because it equals zero — a naive scorecard applies a maximum penalty. The result: healthy companies rated as risky, silently.
The scorecard therefore distinguishes absence from a zero value. An absence carries no penalty: it lowers the confidence level, which is the honest signal. And an entity filing no accounts at all gets an explicit no-conclusion verdict, never a score fabricated out of nothing.
Practical consequence: always read the confidence level before the score. 78/100 at low confidence and 78/100 at high confidence do not justify the same decision.
6,542,232 financial years are in the database, covering 1,586,046 companies, including 175,534 years ended in 2025. Detailed tax-form line items are available across 3,402,583 rows. Worth stating, because it is the real limit of the French market: only a minority of small companies file public accounts, and for the others no vendor can manufacture a financial analysis — we would rather say so than simulate one.
Two routes, two uses. /v1/score/defaillance/{siren} returns a number and its scorecard: that is what you wire into an automated rule. /v1/entreprise/{siren}/sante returns a bilingual read-out — verdict, strengths, warning signs, activity trend, confidence level: that is what you put in front of a human who has to decide.
On the written read-out the internal rule is strict: the model ticks boxes, the code writes the prose. A model fills a closed evaluation grid; every figure, date and sentence is assembled by us, and deterministic guards overrule it when the accounts disagree. No free-form model text is ever served.
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.
Would you rather have a euro invoice and prepaid credits? That is in preparation, and we will not announce a date until it is open. Write to contact@sirenic.eu and we will let you know when it opens.
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
# Client x402 (npm) — le devis 402 est réglé et l'appel rejoué automatiquement
npx x402-fetch https://api.sirenic.eu/v1/entreprise/552032534
# Ou en MCP, dans Claude Code / Cursor
claude mcp add --transport http sirenic https://api.sirenic.eu/mcp| Route | Price | What it returns |
|---|---|---|
/v1/score/defaillance/{siren} (free sample) | $0.10 | Credit risk scoring for a French company: a default-risk score (0-100) at ~12 months from a transparent scorecard — filed financial ratios (structure, profitability, liquidity, net cash, debt service, trend), company age, and a hard BODACC override (open insolvency/liquidation, or closure for insufficiency of assets). Returns the score, a qualitative band, every component with its threshold, and a confidence level. Decision-support indicator — NOT a solvency opinion or credit rating. |
/v1/entreprise/{siren}/sante (free sample) | $0.15 | Company health check for a French company: a bilingual risk read-out — verdict, strengths, warning signs, activity trend and confidence level — from official data only (identity, BODACC alerts, filed financials, sanctions screening). A model fills a closed evaluation grid; every figure, date and sentence is assembled by Sirenic, and guards overrule it when the accounts disagree. Entities filing no accounts get an explicit no-conclusion verdict. For KYB and due diligence. Cached 7 days. |
/v1/entreprise/{siren}/finances (free sample) | $0.01 | Financials of a French company from annual accounts filed at the INPI registry, with Banque de France ratios: revenue, EBITDA, net income, debt ratio, financial autonomy — one entry per fiscal year. Includes the full structured financial statements line items (liasse fiscale tax form, incl. balance sheet) as keyed by INPI, for up to the 20 latest fiscal years. A perimetre_comptable block states whether figures are statutory or consolidated, and flags series where the source conflates the two. |
/v1/secteur/{code_naf}/benchmarks (free sample) | $0.05 | Sector benchmarks and sector statistics for a French NAF activity code (any level: division, groupe, classe, sous-classe): number of active companies, median company age (with quartiles), workforce-bracket distribution, and — when at least 5 companies file public accounts — median revenue, EBITDA margin, pre-tax result and debt ratio. Peer comparison in one call: place a company against its peers. Aggregates only, no personal data; sectors with fewer than 5 companies are excluded. |
The default-risk score is a decision-support indicator. It is neither a solvency opinion nor a credit rating in the regulatory sense, and it guarantees no payment.
Yes. This is a measured difference from scorings served out of a batch-recomputed warehouse: on two companies in a calibration sample, a competitor served a rating built on 2017 and 2023 accounts even though 2024 and 2025 filings existed.
Yes, with adapted handling and a confidence level that says so. This is a deliberate choice: others abstain (“sector not covered”). Abstaining is defensible; we prefer to return an explicitly cautious read-out rather than nothing, and let you judge.
That is your decision and your responsibility. The score is a decision-support indicator, not a solvency opinion: it gives you evidence and its provenance, it does not decide for you and carries no regulated-opinion status.
No account to create, no key to request: the first step is free and shows you exactly what you would be buying.
See a real response (free) Get the quote for /v1/score/defaillance/{siren} Read the route reference Get notified about euro packs