BODACC.io vs API Entreprise 2026: the scopes
Am I allowed to use API Entreprise: who this State service is open to
API Entreprise is a bundle of APIs operated by the Interministerial Digital Directorate (DINUM) on the domain entreprise.api.gouv.fr. Its purpose is to "simplify the administrative procedures of companies and associations that use public services, in particular by relieving them of the need to produce certain supporting documents", and to "facilitate the processing associated with these procedures by allowing the administrations responsible for them to retrieve documents or data produced or held by other administrative authorities" (terms of use, article 1). The legal scope that follows is explicit: the service "is open to administrations within the meaning of article L. 100-3 of the Code of relations between the public and the administration", that is, State administrations, local authorities, their respective public administrative establishments, and public and private law bodies and persons entrusted with a public administrative service mission, including social security bodies. The service's FAQ restates this rule in three lines: API Entreprise is reserved for public actors entrusted with a public service mission; their private contractors "may be recipients of the technical information enabling use of the API but in no case of the data itself"; and the service "is not intended for" individuals, associations or companies. The service's official data.gouv.fr page formalizes this as a grid of eligible audiences: access "Restricted", "Local authority and Administration: Yes", "Company and Association: No", "Individual: No" (API Entreprise page). In other words, a private company cannot obtain this bundle for its own commercial needs, even by funding its use: DINUM "grants authorizations to retrieve information according to the nature of the procedures to be processed (public procurement, public aid, etc.)", and the use cases listed are administrative: processing applications for public procurement, filing and processing grant applications, procedures on one-stop shop portals, fraud detection, form pre-filling. The service itself is presented as a certification tool: "data is taken at the source", the information "is certified and authoritative".
The access path is documented just as precisely, and it explains much of the scope. Applications are submitted on Datapass, the State's processing tool (API Entreprise form), and processing an application takes "on average 11 days depending on the volume of requests"; the FAQ adds that there is no point in contacting support again before 20 days and that no request goes unanswered, with the instructors' dashboard sorting applications by age (processing time). The application is demanding, which is consistent with the nature of the data concerned. A precise legal framework is required: a decree, an order or a legislative reference, knowing that the FAQ explicitly excludes the CRPA, the ESSOC law and the Lemaire law, which "are not sufficient because they indicate a principle of exchange that must be supplemented by a precise legal framework for the intended use". A data protection officer is required, with this remark: "if you do not have a DPO, it is probably because you are not authorized to use API Entreprise". A business contact, a technical contact, the list of data "strictly necessary" with their legal justification, the recipients within the meaning of the CNIL definition and the retention period are also required. The terms of use finally require one application per administrative procedure and one token per application: "the fact for a Partner to use a single token for several different procedures, including by internalizing the management of access rights, results in the immediate revocation of the token". A software publisher does not circumvent the rule: it is up to its public client to submit the authorization application, and the publisher can only ask to be listed on a use case in order to offer a pre-filled form to its clients. Expanding a data scope, changing a contact, transferring an authorization to a colleague or extending a token are all separate procedures, processed one by one by DINUM's services.
The State's limits, measured one by one
This scope is not a flaw: it is an architecture, calibrated for public officials who process procedures, with safeguards proportionate to the data exchanged. The caps and obligations are public, which makes it possible to report them precisely.
| What is measurable | Value recorded on 10/15/2026 | Source |
|---|---|---|
| API sheets in the catalogue | 41, of which 3 announced "coming soon", 1 in beta version, 17 marked "API included by default" | catalogue |
| General cap per IP address | 1,000 requests / minute | developer space |
| Cap per authorization, JSON endpoints | 250 requests / minute, shared among all APIs in the group | developer space |
| Cap per authorization, document endpoints | 50 requests / minute | developer space |
| Tax certificate (DGFiP) | 5 requests / minute | DGFiP sheet |
| Acts and financial statements, beneficial owners, registration certificate (INPI RNE) | 15 requests / minute | INPI sheet |
| Annual and monthly headcounts (GIP-MDS) | 50 requests / minute | GIP-MDS sheet |
| Compliance certificate (Urssaf) | 100 requests / minute | Urssaf sheet |
| Exceeding the cap | HTTP 429, then a 12-hour IP address ban, "fixed and non-revocable duration", all tokens on the account blocked | FAQ |
| Request to increase volume | Refused: "That is not possible!" | FAQ |
| Token lifetime | 18 months, extension reserved for the applicant only, notifications sent 90, 60, 30 then 15 days before expiry | FAQ |
| Availability commitment | 99.5% on the API Entreprise service, and not on the underlying APIs, which depend on the data providers | FAQ |
| Mandatory traceability | Retention of officials' access logs for at least 36 months | Terms of use |
This table also implicitly describes what the State forbids itself from doing. The terms of use recall that the partner must not give access to the data "to the user not benefiting from a sufficient level of authentication", that it must inform the user before any transmission and obtain their explicit consent where applicable, and that the data received may be neither marketed nor communicated to third parties outside the cases provided for by law. On the sanction side, the mechanism is described without ambiguity: in the event of failure to take into account 429 codes or exceeding the overall cap, "your IP is temporarily banned from our servers for a fixed and non-revocable period of 12 hours", all tokens being blocked during that time, and calls from a banned IP returning "no HTTP codes": the server simply does not respond. The FAQ also specifies that no exception exists: to the question "I need more volume, how can I be whitelisted?", the answer is "that is not possible! The volume indicated is the maximum possible so that we can guarantee the quality of service to all our users". Three technical constraints complete the system: calling from a website in the front end is prohibited, because the token would then be disclosed; CORS is not allowed; the API is therefore consumed from a server, never from a browser (technical prerequisites). The exact cap of an endpoint is returned in the RateLimit-Limit, RateLimit-Remaining and RateLimit-Reset headers, with a Retry-after in seconds on 429 errors. One exception exists for developers: the routes for monitoring the status of providers have "no rate restriction" and "do not consume your quotas". Finally, the catalogue exposes explicit error codes, including "XX005: unavailable for legal reasons, this entity is not accessible by any organization": the door can therefore be closed not only by the cap, but also by law.
Free public data: what it really covers (and what it leaves out)
When you are not a government body, free access does exist, but you need to know exactly what it contains, because the word "open" covers very different scopes. The Insee Sirene open data API is the broadest of the free gateways: it provides legal units and establishments registered in the Sirene directory since its creation in 1973, including closed units, with single, multi-criteria and phonetic search and historical data. Two complementary services are worth knowing: the "informations" service, which gives the date the data was last updated, and the "liens de succession" service, which returns the predecessors and successors of an establishment, that is, the continuity of a business after a sale or takeover. Access is described as "open with account", the cap is 30 requests per minute and the announced availability is 99.5%. Two limits are set by Insee itself, and they matter to anyone who wants to reuse this data in a product. First, persons registered under partial diffusion, in accordance with article R123-232-1 of the French commercial code, must not have their information "either redistributed in full or used for prospecting purposes" — in other words, the diffusion status must be checked and respected on every use. Second, the former nonDiffusibles service no longer exists: since 21 March 2023, "the non-diffusible 'N' modality no longer exists", the status has been converted to partial diffusion "P", and "there are therefore no more non-diffusible 'N' units in the Sirene database". The variant that includes units under partial diffusion is under restricted access, still at 30 calls per minute. And if you prefer files to API calls, the full database can be downloaded from data.gouv.fr: the dataset page, however, states a monthly frequency, with the last update observed on 1 October 2026 (Sirene database).
The second gateway is the API Recherche d'Entreprises, the one that powers the company directory, the general-public lookup site. It is open without an account, accepts both text and geographic search, and its limitations are acknowledged on its page: "because the API is completely open access, it has limitations". Not accessible are the predecessors and successors of an establishment, non-diffusible companies, and companies whose registration in the RCS was refused. The page adds this decisive point: this API "does not provide access to the complete data of the Sirene database, but only allows searching for a company by its name or address". It returns the name, SIREN, SIRET and NAF code, with filters by NAF code, postal code, employee headcount bracket or sole proprietorships. On throughput, the technical specification states 7 requests per second per IP address, 30 per second per ASN, with two explicit caveats: "this value corresponds to a maximum limit and does not constitute a guaranteed throughput", and "in the event of excessive use [...] priority is then given to internal systems and public services". The ecosystem around this API is documented: it is backed by four associated datasets, including the Ademe's list of RGE-certified companies, live performance entrepreneur licences and licence applications, the national directory of elected officials and the Sirene database itself. So here, for a general-public product, is the essentials of the identity reference base: accessible, documented, but rate-limited and deliberately restricted to the data strictly necessary for a search. It is a good starting foundation, provided you accept caching, ingesting monthly files and working with partial scopes. Note that this is the route the State itself designates as the general-public alternative to API Entreprise, and that the Sirene database is published under the Licence Ouverte: reuse, including commercial reuse, is therefore provided for by the licence, subject to respecting partial diffusion statuses.
What this free foundation does not contain, however, is the dated movement of companies. Neither legal announcements, nor insolvency proceedings with their timeline, nor account filings as dated events, nor sales and transfers of businesses. The identity is there, the official reference of registered companies is there; the narrative of decisions is not. And this absence is not an oversight: it corresponds to another source, published separately, that of the BODACC. Serious monitoring therefore requires ingesting this stream at the pace of its five weekly publications, then treating each notice as a dated event, not as a company attribute. To measure it, we queried the public endpoints ourselves on 15 October 2026, and the finding comes down to three orders of magnitude. A cap of 30 requests per minute, for an entire service and not per user, imposes a cache-based architecture and file-based ingestion as soon as traffic becomes real: this is not a limit to endure, it is a design constraint to build in from day one. The monthly frequency of the Sirene files means that fine-grained monitoring of creations, deregistrations and head office transfers cannot rely on that channel alone. And partial diffusion status must be respected on every use, including in automated processing. These are design constraints, not criticisms: they simply map out the place of each source in a value chain. A monitoring product must therefore build its own temporal index: no national endpoint directly answers the question "what has changed at this company in the last six months?". You have to ingest the movement, date it, attach it to the SIREN, then reconstruct the timeline, which we expose as structured fields (notice family, court, department, city, notice identifier, SIREN) in the stream of published announcements.
| Free access | What it gives you | What is missing | Cap |
|---|---|---|---|
| API Sirene open data (Insee) | Legal units and establishments since 1973, closed units, historisation, succession links | Legal announcements, accounts, proceedings, directors in stream form | 30 requests / minute |
| API Sirene with partial diffusion | Same scope, partial diffusion included | Restricted access, prospecting use prohibited | 30 calls / minute |
| Sirene database (files) | Complete files and stable URLs | Monthly update, ingestion on your side | None (download) |
| API Recherche d'Entreprises | Text and geographic search, name, SIREN, SIRET, NAF | Non-diffusible companies, successions, registration refusals, full record | 7 requests / second |
| API BODACC (DILA) | The full text of commercial announcements — 51,043,433 records as of 15/10/2026 | Linking to companies, alerts, timelines | Not published by the publisher |
Company activity: BODACC, what the Entreprise API catalogue does not contain
The Entreprise API catalogue lists, as of 15 October 2026, 41 API records fed by Insee, INPI, Infogreffe, DGFiP, Banque de France, Urssaf, MSA, GIP-MDS, Douanes, DJEPVA, SIAF, Ademe, Qualibat, Qualifelec, Opqibi, France compétences, ProBTP, CIBTP, CNETP and FNTP. It contains companies' civil status, their addresses and registered offices, their corporate officers, their annual and monthly headcounts, their turnover, their tax returns, their capital links, their balance sheets, their certifications and a series of compliance certificates. BODACC is not there: none of the 41 records covers legal notices, nor the publication stream of commercial court decisions. This point deserves to be highlighted, because it defines a boundary: queries are always made on an entity — a SIREN, a SIRET, an RNA or EORI number — never on an event. There is no way to ask "what was published this week?" or "which companies opened insolvency proceedings in this department?". Company activity is therefore not absent from French open data — it is elsewhere, and you need to know that before building a product on it. The FAQ also explains why some APIs are said to be "included by default": those that only deliver public information are not subject to any additional legal framework and are therefore attached to all access granted. Even in this favourable case, the data remains reachable through the main door, the bundle reserved for administrations and their technical providers. A detail in the catalogue confirms this reading: among the APIs marked "included by default" are the RCS extract, acts and balance sheets, legal unit data in open data, and intra-community VAT numbers — information that is mostly public, but distributed through the same single door, subject to authorisation.
This "elsewhere" is the DILA, the publisher of the Bulletin officiel des annonces civiles et commerciales. The BODACC dataset, governed by article R 123-209 of the French commercial code, publishes the notices issued by commercial court registries, civil courts with commercial jurisdiction and insolvency practitioners: sales and transfers, registrations, establishment creations, modifications and removals of natural or legal persons, insolvency proceedings, conciliation proceedings, professional reinstatements, miscellaneous notices and notices of filing of company accounts. The bulletins are published five times a week — Monday-Tuesday, Wednesday, Thursday, Friday, Saturday-Sunday — and are divided into three series: BODACC A (sales, transfers, registrations, creations, insolvency proceedings), BODACC B (general modifications and removals) and BODACC C (annual accounts filings). Data access is free via HTTPS and FTPS (streams and archives), and a BODACC API can be queried without a key. We measured both databases on 15 October 2026: the DILA API returns 51,043,433 notices, with the latest publication dated 14 October 2026 (publication 20260197); our own API returns 51,029,854 for the same latest publication date. Both services are aligned on the most recent edition, the volume difference being due to the history retained on each side. A concrete example, taken from the latest publication: notice C202601972306, an initial accounts filing notice published on 14 October 2026, issued by the registry of the mixed commercial court of Saint-Pierre and linked to SIREN 988208435. The structure exposed by this API is stable and documented: notice identifier, publication number and date, notice type and family, court, registration registry, department, city, and a field dedicated to each type of notice — judgment, act, general modifications, removal from the RCS, filing, previous operators or owners. The dataset, created in 2015, is published under the Open Licence and showed a last update on 14 October 2026; the publisher, attached to the services of the Prime Minister, remains reachable by email for any reuse question.
One nuance of timing remains, documented by the publisher itself. The official BODACC API record states that the digital data represent "the full text of the notices" and that "a weekly update of the data is in place". The XML streams, for their part, are deposited in a folder whose last modification was dated 14 October 2026. The official data is therefore fully available, in two complementary forms: a stream to ingest publication by publication, or a database queried at its own pace. Two caveats accompany this availability, and they are written by the DILA. Civil notices — personal reinstatements, estates accepted up to the amount of net assets — are not included in the database, the publisher citing the protection of the rights of over-indebted persons and heirs. And "the provision by the DILA of datasets that may contain personal data does not exempt the reuser from complying with the French Data Protection Act", with the obligation to take into account update requests published in the dataset's discussion forum. Alongside this, the documents themselves — acts and balance sheets — are distributed by the INPI "Acts and balance sheets" API on the RNE, whose record announces a daily update: "as soon as a new document is available concerning a company, this document is available during the day and is added to the list of acts and balance sheets of the company concerned". These are documents, accessible only to authorised organisations, and not events attached to a company's timeline. The difference is not one of quality, it is one of nature: the State publishes the source and its documents, we recompose the thread. This cadence — daily for documents, weekly for the queried database — is not a shortcoming, it is the pace of an official publisher; it simply has a practical consequence for a monitoring tool, which must then work on the streams to stay as close as possible to publication.
| Company activity | What the State makes available | What BODACC.io makes available |
|---|---|---|
| Legal notices (series A, B, C) | XML streams, archives, BODACC API (DILA), five publications per week | The dated notice feed, in French and eventually in 20 languages |
| Latest publication observed | 14/10/2026, publication 20260197 — 51,043,433 notices | 14/10/2026, publication 20260197 — 51,029,854 notices |
| Documents and acts | INPI RNE "Acts and balance sheets" API, 15 requests/minute, authorised organisations | Indexing of accounts filings as dated events |
| Notice ↔ company linking | To be built by the reuser | Native: SIREN, SIRET, registry, department, city |
| Alerts and tracking over time | Not provided by the publication | Alerts and history by company and by person |
Building a consumer product: API Entreprise or BODACC open data?
If your project is aimed at the general public — a search site, a business monitoring tool, an analytics product — API Entreprise is not a building block you can use, and this is not an obstacle to work around but a framework to understand. The eligibility grid in its official listing marks "No" for both "Company and Association" and "Individual", and the terms of use reserve the service for administrations and their technical providers, with the latter never having access to the data itself. This restriction is not an arbitrary closure: the service exchanges administrative documents protected by confidentiality, and it reserves them for organizations that need them to process a case — this is the logic known as "Tell us once", which aims to avoid asking a company again for a supporting document that an administration already holds. The State does not leave anyone without a solution, moreover: the same listing directs non-eligible users to its own freely accessible alternatives, namely the business directory, the Sirene API produced by Insee and the RGE professionals API produced by Ademe. The distinction is clear and worth keeping in mind: the State opens its identity registry widely, and reserves the exchange of administrative documents for authorized organizations. A consumer product therefore lives on the first half of this equation, and must organize itself accordingly: more open data to assemble yourself, fewer certified documents to display. A business monitoring product therefore does not need to present itself as a public service: it simply needs to stay within the scope of open data, where reuse is provided for by the license. The FAQ also describes two uses of the service: obtaining data in back office by an authorized agent, and pre-filling a form for the public. Even this second use, the closest to the general public, remains reserved for a public service: the model is consistent from start to finish.
The workable path consists of three documented building blocks, in this order. Identity first: Sirene for legal units and establishments, with a cap of 30 requests per minute per account, which mechanically pushes you to cache results and ingest monthly files rather than querying the source continuously. Search next: the API Recherche d'Entreprises, at 7 requests per second per IP, very well suited to autocomplete, but which only returns name, SIREN, SIRET and NAF code and excludes non-diffusible entities such as successions. Movement finally: the BODACC feeds from DILA, to be ingested publication by publication, with this obligation recalled by the publisher: "the reuser undertakes to take into account requests to update published data made occasionally by DILA in the discussion forum of the dataset on data.gouv.fr". To this mechanics is added the work that no one does for you: matching each announcement to an identified legal unit, tracking the chronology of a procedure over time, detecting changes in a company's status, and decoding a judgment without mistaking the series or the type of notice. Each of these steps has a development cost, then a maintenance cost: the quality of a monitoring product depends less on access to data than on keeping this chain running over time. The expected practices are also documented: send an explicit User-Agent header, respect the 429 code and the Retry-After header rather than insisting, and prefer bulk downloads to repeated queries. At this stage, the assemblable chain covers identity, search and movement, but leaves aside financial documents and certificates: these go through INPI, court registries or the administration, and remain out of reach for a consumer product.
This is precisely what BODACC.io does, and we embrace it as a different choice, not as a superiority. We start from the same public sources, we recompose them into navigable objects — announcements, companies, directors and natural persons — and we expose them without an application or authorization. Concretely: consultation is free and requires no registration, a company can be found in the business directory, a person can be tracked with their mandates in the dedicated space for directors and natural persons, each new publication can trigger an alert, and the data API is public. The same content is exposed in French and, eventually, in twenty languages, and access to consultation requires neither an account nor justification of use. The limits announced on our side are 60 requests per minute and 10,000 requests per day without a key, 600 per minute and 100,000 per day with a key, with an edge cache of 300 seconds on common searches, so that repeated requests do not consume the quota. For consumer use or a product integration, these thresholds compare to the 30 requests per minute of Sirene, the 7 requests per second of the business search and the 250 requests per minute per authorization of API Entreprise. The guaranteed throughput is lower on the State side because the target is not the same, and because a public service must guarantee quality of service to all its users rather than to a single integrator. On our side, we publish announcements with their useful attributes — type of notice, family, court registry, department, city, company identifiers — and we link them to company and person profiles, which allows searching by event as well as by entity. None of these functions depends on an authorization: they follow solely from the fact that the source data is open and redistributable.
What it really costs, in practice
The first line is simple: use of the API Entreprise "is free" and "investment and operating costs are covered by DINUM" (FAQ, how much does the API Entreprise cost). But the same answer immediately adds this sentence: "the costs of connecting to API Entreprise are yours." The real cost is therefore not a fee, it is a project, and it is measured in engineering hours and compliance obligations. You need an IT department or a software vendor capable of integrating the API, a back-end application since calls from a browser are prohibited, token management with an 18-month lifespan and mandatory extension at the sole initiative of the requester, access logging kept for 36 months, and a degraded mode planned for incidents — the FAQ insists: "Dites-le-nous une fois must not block users in the event of technical incidents." You also need to size timeouts: 5 seconds recommended for a JSON response, 12 seconds for a PDF or a ZIP, with the API itself waiting a maximum of 10 seconds for its providers before returning a 504 error. Add to that an authorization file for each administrative procedure, with legal framework, DPO and description of recipients, which must be updated as soon as a contact, a scope or a use changes. For a software vendor, you also need to collect and rotate a token per client administration, or go through the delegation mechanism described in the developer space; the official SDKs cover Ruby and Node.js, with Python, PHP and Java implementations announced as coming soon. These are real costs, but predictable, and they are the counterpart of a service that commits the State to certifying the data exchanged. These are organizational costs, not license costs, and they are amortized over time as long as the requested data scope remains stable.
What the State offers in return deserves to be said, because it weighs in the balance: free technical and functional assistance, an office open by videoconference twice a month, status pages per provider, monitoring routes with no call limit, a test environment to evaluate the service before any access request, and a 99.5% availability commitment on the service. Add to that public information on usage and availability, including statistics consultable without an account. For a public body processing files, this combination — certified data, enforceable documents, assistance — is hard to match. The open data path, for its part, has a different cost structure: nothing to pay, everything to build. Downloading BODACC feeds costs nothing, but you have to ingest them, store them, reconcile them with Sirene, account for update requests published in the forums, then maintain the chain over time. The Sirene API requires an account and key management, with 30 requests per minute for an entire service — a ceiling rarely sufficient for mass-market use without cache or monthly files. At our end, consultation is free and without registration, the data API is open without a key, and public ceilings are simply exceeded by creating an account: 60 requests per minute and 10,000 per day without authentication, 600 and 100,000 with a key, with details in our offers. The right way to compare the two paths is therefore to look at the maintenance burden: ingesting a daily feed, complying with a dissemination status that may change, maintaining a consistent cache and managing a rotating key — four tasks that the State handles for its partners, and that an open data reuser assumes alone.
The real question is therefore not "who is cheaper" — both paths are free — but "who bears the burden of integration, refreshing and formatting." On the API Entreprise side, this burden is with the administration or its vendor, in exchange for certified data, taken at the source, which is authoritative. On the open data side, the burden is with the reuser: feed ingestion, reconciliation of announcements with companies, cache, languages, ergonomics. It is this recomposition work that we handle on our side, without claiming to replace the source, and that explains our own ceilings, broader than those of the statistical register because our use is narrower. Three figures honestly summarize the comparison: 30 requests per minute for an entire service at Insee, 250 requests per minute per authorization on the API Entreprise side for a public body, 60 requests per minute and 10,000 per day at our end without even a key. None of these figures is better in absolute terms: they reflect three different missions — a national statistical register, an exchange service between administrations, and a monitoring tool open to all. Each has its ceilings for good reasons, and a serious integrator starts by reading them before sizing its architecture. We publish ours openly, rather than letting a developer discover them at the first overrun. A difference in framework finally deserves to be noted, because it weighs on business models: the API Entreprise terms of use prohibit commercializing the data received through the service, while the Open License of BODACC and Sirene explicitly authorizes reuse, including commercial. This is what makes a product like ours possible, and it is also why these two worlds do not compete head-on.
| Cost item | API Entreprise (State) | Open data (reuser) | BODACC.io |
|---|---|---|---|
| Data fee | €0, covered by DINUM | €0 | €0 |
| File and authorization | One request per procedure, processing ~11 days | Not required | Not required |
| Technical integration | Your responsibility (IT department or vendor, back-end mandatory) | Your responsibility (ingestion, reconciliation, cache) | Direct HTTP call, open API without key |
| Guaranteed throughput | 1,000 req/min per IP, 250 per authorization (JSON), 50 for documents | 30 req/min (Sirene), 7 req/s (Recherche d'Entreprises) | 60 req/min and 10,000/day without key; 600 and 100,000 with key |
| Refresh | Daily for INPI documents, weekly for the DILA BODACC database | Monthly for Sirene files | At each BODACC publication |
| Assistance | Free, bi-monthly open office, status pages | None | Support by account |
What the State does better than we do
It must be said plainly: when it comes to French company data, the State is the primary source. It produces Sirene, publishes the BODACC, holds the certificates, tax returns and registers, and stakes its responsibility on certifying that information. Its promise is stronger than ours on one decisive point: on its site, "data is taken at the source," and "the information is certified and authoritative" (API Entreprise home page). We, by contrast, reassemble this material: we certify nothing, we issue no legally enforceable document, and we do not claim to be an authority. For an administration processing a public contract, a grant application or an audit, the answer is simple and we give it as it is: the tool designed for that is API Entreprise, it is free, it is proven, and it is calibrated for that use. In that context, our service is not meant to act as a source of proof, and it will not. Certification, authentication of a document and legal enforceability are the responsibility of the State and the court registries: these are responsibilities that cannot be delegated to an analysis tool. On the substance, there is in fact no possible debate: administrative data has legal value only if the administration that produces it authenticates it. We say this all the more readily because our data comes from there: without the DILA's publication of the BODACC, without the registers kept by Insee and the INPI, we would have nothing to analyze. Our work is a work of reading and connecting, not of producing information. And if your use case is processing a file or verifying a supporting document, stop here: API Entreprise is the tool designed for it, it is free, and it delivers documents that are authoritative.
The list of what the State holds and we do not reproduce is long, and it deserves to be cited precisely: tax and vigilance certificates, tax returns, shareholding links declared in CERFA 2059F and 2059G, revenue for the last three fiscal years, balance sheets held by the Banque de France, annual and monthly headcounts from Urssaf and the MSA, MSA, ProBTP, CIBTP and CNETP contribution compliance, public works professional card, customs EORI registration, RGE, Qualibat, Qualifelec, OPQIBI, Qualiopi and Bio certifications, association subsidies, foundation data, and the extract from the trade and companies register annotated by the court clerks. The catalog also specifies that the most sensitive data, such as beneficial owners, is reserved for use by an "authorized agent or authenticated user": access is backed by a legal framework, not by a simple sign-up. We do not have that, and we will not: this data is protected by confidentiality rules and intended for supervised administrative uses, which is legitimate. To this add two points of honesty measured on October 15, 2026. First, the DILA's BODACC API returns 51,043,433 announcements while ours returns 51,029,854: the volume gap is not in our favor, and we do not claim to be the reference database. Second, the State publishes its own usage figures, including a public statistics page (API Entreprise statistics) and per-provider availability pages, whereas our communication remains more modest. Finally, the DILA provides supporting documentation, templates and a long history: our added value rests on processing, not on documentary authority. We display neither a Kbis, nor a vigilance certificate, nor a downloadable registration extract: these documents are obtained at the source, from the court registry, the INPI or the administration, and it would be misleading to suggest that a summary sheet replaces them.
What we bring is different, not superior. The State documents a company at a given moment and certifies its documents; we re-read the flow of its decisions to turn it into a readable narrative: an announcement as soon as it is published, a company with its history, a person with their mandates and proceedings, an alert when something changes, and, for a given notice, a direct reading of what it implies — for example the duration of a measure ordered against a director, calculated from the notice and displayed as a key data point. The opening of insolvency proceedings, for example, appears in the official feed as one notice among others; we turn it into a readable status for the company concerned, with its date and its scope. This type of use — securing a receivable before delivery by checking a client's situation — is a commercial monitoring use, not an administrative processing use. Both exist and do not answer the same question: one says "this document is authentic and authoritative," the other says "here is what has just changed for this company, and since when." If you are an administration, the State is your right interlocutor, and we say so without reservation: our sources are its data, and our work is worth something only because its work exists. If you track companies in order to decide — grant credit, sign a contract, source a supplier, monitor a competitor — that is our territory, and it is played out on the same public data, better connected. This layer of interpretation is our only contribution, and it remains entirely dependent on the quality of official publications. That is the only promise we make: to make readable, alertable and translatable an official flow that, without processing, remains a series of raw notices and records.
Our sources, verifiable line by line
All the facts on this page were collected on October 15, 2026. Each link points to the official page that carries the information.
- Eligibility, free access, timelines, quotas, bans, partner commitments: API Entreprise FAQ and terms of use
- Volume caps, technical prerequisites, SDK, publisher delegation: developer space
- Composition of the bundle, providers, "included by default" badges: API catalog
- Restricted access, eligible audiences, open access alternatives: API Entreprise page on data.gouv.fr
- Authorization form and review: Datapass
- Quotas recorded page by page: tax certificate, acts and financial statements, GIP-MDS headcounts, compliance certificate, RCS extract
- Sirene directory: scope, partial dissemination, 30 requests per minute — Sirene open data API and Sirene API with partially disseminated units
- Monthly update of Sirene files: Sirene database on data.gouv.fr
- Company search: scope, 7 requests per second — API Recherche d'Entreprises page and technical specification
- BODACC publication, series A/B/C, five publications per week, open license: BODACC dataset, DILA feeds and archives, BODACC API page
- Volume and freshness measurements: DILA BODACC API — 51,043,433 notices, latest publication on 10/14/2026 — and BODACC.io public API — 51,029,854 notices, same latest publication
- Usage limits of our own API (60 requests per minute and 10,000 per day without a key; 600 and 100,000 with a key): BODACC.io API documentation