Coming soon
The OCR API is not available yet
This page describes an interface that is being designed. There is no endpoint accepting requests, no key to obtain and no timetable being promised. It exists so that anyone planning an integration can see the intended shape early.
This product is not operational. Nothing on this page can be used yet, and the page is excluded from search results until it is.
What OCR API does
The API is intended for teams that already have a system receiving invoices — an ERP, a procurement tool, an internal dashboard — and would rather have extraction happen there than in a browser tab.
Nothing described below is live. The endpoints do not exist, no credentials are issued, and no request will succeed. Treat this page as a design note, not as documentation for something you can call today.
It is published early for one reason: an integrator deciding between vendors deserves to know what is planned and what is not, rather than discovering the gap after committing.
What gets extracted
Planned operations
- Submit a document for extraction
- Check the processing status of a submission
- Retrieve the structured result
- List previously submitted documents
- Delete a document and its record
Planned response contents
- Document metadata and status
- Invoice-level fields with a confidence value
- A line-item array
- Detected currency and locale hints
- Page count
How it works in practice
Not accepting requests
There is no live endpoint. Any call you make today will fail, because nothing is listening.
No keys are issued
There is no key management, no developer account area and no way to authenticate. API access is not part of any current plan.
Design goals, not features
Idempotent submission, a stable machine-readable error envelope and explicit rate-limit headers are what the interface is being designed around. None of them is implemented.
No SDK and no webhooks
Neither exists. They are absent from this page rather than described as though they worked.
Who it is for
Product teams
Adding document capture to an existing application.
Internal tooling
Automating an inbox of supplier invoices into a database.
Integrators
Connecting extraction to an accounting package on a client's behalf.
Supported formats
- The planned interface targets the same inputs as the browser tools: PDF, JPG, PNG and WebP.
- No upload endpoint is live, so no size or rate limit is in force.
How your documents are handled
- The browser tools and this API are not the same thing where your documents are concerned. In the browser, recognition happens on your device and the file is not uploaded. An API works the other way round: a document sent to an endpoint has necessarily left your machine.
- That difference is the reason this page does not repeat the browser's privacy wording. If the API is built, its handling of documents will be described on its own terms, and only once it is true.
- No retention period, encryption guarantee, processing location or compliance certification is being claimed, because none has been built or audited. A page promising those before they exist would be worth less than one that says nothing.
- If keeping documents on your own device matters to you, the browser tools already do that today and will continue to.
Full detail, including what we deliberately do not claim, is on the Security page.
Frequently asked questions
Ready to try OCR API?
The API accepts no requests and there is no date to give you. Tell us what you would build, how many documents a month, and in which languages — that is what decides whether it gets built and what it does first. Asking does not reserve a place. If you need extraction today, the browser tools already do it: /en/invoice-ocr