Is there a free API tier?

No — free-tier API keys get scraped and abused, so the free tools are web-only. The cheapest real path is a one-time $15 credit pack, which unlocks keys with no subscription.

What are the rate and size limits?

Plan-dependent file sizes (25 MB to 250 MB) and monthly page allowances, with a 60-requests-per-minute API rate limit. All published on the docs page, all enforced server-side, all returned as machine-readable errors.

No, deliberately — the API is two endpoints with multipart uploads and JSON responses. The docs show working curl, Node and Python that you can paste. An SDK would be a wrapper around fetch.

How stable is the response schema?

Extraction output is schema-bound per document type — the model cannot add or rename fields. Additions arrive as new nullable fields, announced in the changelog.

Last updated 2026-08-03

Solutions · Developers

The OCR API

Send a document, get text and structured data back. Two calls, no SDK required, and every sample in the docs came from a live request.

Read the API docs Plans from $19

You need document text inside your product, and the hyperscaler OCR products come wrapped in a cloud console, an IAM model and a pricing calculator. You want an endpoint, a key, and numbers you can trust.

This API is deliberately small: OCR in two calls (submit, poll), structured extraction in one synchronous call, plain error bodies with machine-readable codes, and documentation whose every sample was captured from a real request — because a wrong sample in docs ships into someone else’s product.

The numbers a developer actually asks

Latency, languages, auth and limits — measured and stated, not marketed.

Measured 2026-08-03

Speed: OCR reads a clean page in 0.5–0.65 seconds on our measurements across formats; structured extraction answers synchronously in 2–5 seconds per document. A 40-page PDF runs as a queued job you poll, so nothing holds a connection open for minutes.
Languages: 161 recognition languages, and unlike most vendors we publish per-script measurements — including the imperfect ones. Sinhala 97.5%, Tamil 95.4%, Arabic 93.8%, Korean 80% with the failure mode explained.
Keys that cannot leak from us: We store only the SHA-256 of your key, under that hash — shown once at creation, unrecoverable by support, revoked within 60 seconds. A first credit pack of $15+ unlocks keys with no subscription at all.
Failure is machine-readable: Limits return 402/429 with a code and a top-up URL in the body, plus Retry-After where it applies. Your integration can distinguish "out of credits" from "broken" without parsing prose.
Where your documents live: Processed on our own servers in Frankfurt, Germany. Originals are deleted within an hour of processing, results within 30 days, and nothing is ever used to train models — ours or anyone else’s. The security page and DPA spell out the rest.

Method and raw results: how we measure. Every figure is dated and reproducible.

How you use it

OCR: POST /v1/ocr with the file → 202 and a job id → GET /v1/jobs/{id} until 200. Full text, confidence, and every word with its position on the page.
Extraction: POST /v1/extract with the file and a type (invoice, receipt, purchase_order, delivery_note, goods_received_note) → schema-bound JSON in the response, with arithmetic checks included.
Start today: Any paid plan includes the API; pay-as-you-go developers can buy a single $15 pack instead. Keys live in your dashboard; the docs carry curl, Node and Python.

What to know before you commit

The limits, stated here rather than discovered in your evaluation.

KNOWN LIMIT

No webhooks yet — polling is the completion signal. Webhooks on job completion are on the roadmap and marked "launching" on pricing, which is our word for "not shipped". Until they exist, poll the job endpoint; the backoff pattern in the docs keeps you inside the rate limit.

KNOWN LIMIT

One region. Everything runs in Frankfurt. If your compliance requires processing outside the EU, we are not your vendor today — and we would rather say so here than in a sales call.

Questions buyers ask

Is there a free API tier?: No — free-tier API keys get scraped and abused, so the free tools are web-only. The cheapest real path is a one-time $15 credit pack, which unlocks keys with no subscription.
What are the rate and size limits?: Plan-dependent file sizes (25 MB to 250 MB) and monthly page allowances, with a 60-requests-per-minute API rate limit. All published on the docs page, all enforced server-side, all returned as machine-readable errors.
Do you have SDKs?: No, deliberately — the API is two endpoints with multipart uploads and JSON responses. The docs show working curl, Node and Python that you can paste. An SDK would be a wrapper around fetch.
How stable is the response schema?: Extraction output is schema-bound per document type — the model cannot add or rename fields. Additions arrive as new nullable fields, announced in the changelog.

Read the API docs Plans and credits

Prefer a conversation first? Write to the people who built it