OCR for AI agents

Turn content into AI-ready markdown

Powered by Mistral OCR 4, ContextOCR converts PDFs, images, and .eml files into markdown with tables, page breaks, image descriptions, decoded QR code and barcode values, and attached files from .eml uploads.

Test your file Get API key

2-4 sec

for most files

100

free credits

per 1,000 credits

One endpoint, one call per file.

Send the file. Get back markdown.

Request

curl -X POST "https://api.contextocr.dev/v1/ocr" \
  -H "Authorization: Bearer ak_live_..." \
  -H "X-Request-ID: claim-1042" \
  -F "[email protected]"

Response

{
  "markdown": "# claim-packet.pdf\n\n### Page 1\n\n| Field | Value |\n| --- | --- |\n| Claim ID | CLM-1042 |\n| Amount | $1,284.20 |\n\n- Signature present below approval block.\n- QR_CODE: CLM-1042-APPROVED",
  "request_id": "claim-1042",
  "usage": {
    "pages": 2,
    "charged_credits": 2,
    "remaining_credits": 498
  }
}

Where it helps

Use it when plain text is not enough.

Built for documents and emails where the layout, images, tables, and decoded codes matter as much as the text.

AI agents

Give agents clean markdown with page context, image notes, and decoded QR code and barcode values they can act on.

Try Mistral OCR 4

Try Mistral OCR 4 on your own PDFs, images, and emails, then get clean markdown with page context and decoded QR/barcode values.

Email processing

Process .eml files the way customers saw them, including attachments, HTML layouts, embedded images, and QR codes that plain-text parsing misses.

Why teams switch

Replace custom OCR infrastructure with one API.

ContextOCR uses Mistral OCR 4 for extraction and returns clean markdown that keeps the surrounding context: file structure, .eml bodies and attachments, decoded QR/barcode values, image descriptions, tables, and page breaks.

Try the preview Read docs

Tables

Structured tables are preserved as markdown.

Page breaks

Page boundaries stay intact for citations and review.

QR codes and barcodes

Decoded values are returned in the markdown output.

Image descriptions

Logos, screenshots, stamps, signatures, and other images are described in text.

Pricing

Start small. Pay for pages.

Try real files first. When you are ready to use the API, credits are based on processed pages instead of seats or subscriptions.

$9 per 1,000 credits

Start with 100 free credits. Each processed page uses one credit.

Get API key Try the preview

Each PDF page, image, rendered .eml page, and page from a file attached to an .eml upload counts as one credit. Enable auto top-up when you are ready for production volume.

Integration details

Drop document OCR into your app fast.

One API call gives you structured output, job traceability, usage details, and no file storage to clean up.

One upload endpoint

Send PDFs, images, and .eml files to the same API instead of stitching together separate parsers.

Useful JSON back

Get markdown, page context, decoded QR and barcode values, usage, and remaining credits in one response.

Trace every job

Pass your own X-Request-ID and match each OCR result to your logs, workflows, and customer records.

No storage cleanup

Files and OCR results are not stored after processing, so ContextOCR does not become another document store to manage.

FAQ

Is this right for you?

Answers to the questions teams usually ask before integrating.

What does ContextOCR add on top of Mistral OCR 4?

ContextOCR turns Mistral OCR 4 output into production-ready markdown your app can use immediately: page boundaries, tables, image notes, decoded QR and barcode values, and .eml bodies plus attachments in one clean response.

Will it work on messy real-world files?

Yes. It is built for invoices, receipts, shipping labels, screenshots, support tickets, scanned forms, product packaging, and customer emails where layout, visuals, and hidden codes matter.

How fast can my team ship with it?

Start with the preview, verify the markdown on your own files, then integrate one upload endpoint. Pass your own request ID for traceability and get clean JSON back for your workflow.

What happens to uploaded files?

Files are processed for OCR and are not stored after processing. You get the structured result you need without turning ContextOCR into another document repository to manage.

Try it first

Try it on one of your files.

Upload a short file, see how Mistral OCR 4-powered extraction looks in ContextOCR markdown, and use the API when the output fits your use case.

Try for free View examples

{ "markdown": "# claim-packet.pdf\n\n### Page 1\n\n| Field | Value |\n| --- | --- |\n| Claim ID | CLM-1042 |\n| Amount | $1,284.20 |\n\n- Signature present below approval block.\n- QR_CODE: CLM-1042-APPROVED", "request_id": "claim-1042", "usage": { "pages": 2, "charged_credits": 2, "remaining_credits": 498 } }