Meet Atlas Live: liveness detection with no gesture required
Atlas

Documents Available

Atlas Read

Extract data from any document in Arabic, French or English with OCR and vision-language models.

Request a demo Start in the sandbox

Access by invitation: no account yet? Ask for your sandbox workspace.

What it checks

Any document, read in its own language.

Atlas Read recognises the document, extracts the fields you ask for and gives its confidence for each one.

  • Classification

    The document type is recognised, or you declare it on upload to go faster.

  • Arabic, French, English

    Latin and Arabic scripts, bilingual documents included: each field is read as it is written.

  • OCR and vision-language models

    The layout is taken into account, not just the text: each value is tied to its field.

  • Confidence per field

    Each value comes with a confidence level; below your threshold, the document goes to review.

  • Business validations

    Bank number check digits, plausible dates, consistent amounts.

  • Logged corrections

    A correction made in review keeps its author, its time and the original value.

How it works

Send, read, check, deliver.

  1. 1 You send the document PDF, image or phone photo, one or several pages, through the API or from the console.
  2. 2 Atlas classifies and reads it Type recognised, fields extracted with their confidence.
  3. 3 Doubtful cases go to review An analyst corrects or confirms; the correction is logged.
  4. 4 The result is delivered A signed webhook on every status change, or a read through the API.
  1. 1 Classify The document type is recognised
  2. 2 Read Fields extracted in French and Arabic
  3. 3 Check Confidence per field, doubtful cases reviewed
  4. 4 Deliver Signed webhook to your system
Specimen: synthetic document.

For developers

One document, structured fields.

Declare the type and the fields you want, or let Atlas classify the document. The result arrives by signed webhook.

Developer guide Start in the sandbox

Request

POST /v1/documents
Authorization: Bearer atlas_sk_sandbox_…
Content-Type: multipart/form-data

type=proof_of_address
fields=holder_name,address,issue_date
file=@proof.pdf

Response 200 OK 

{
  "id": "1a7fcc35-…",
  "status": "done",
  "type": "proof_of_address",
  "fields": {
    "holder_name": { "value": "Amina Haddad" },
    "address": { "value": "12 Acacia Avenue" }
  }
}

Take the next step

Try Atlas Read on your own flows.

A demo on your real cases, then a sandbox space with synthetic identities to integrate without any real data.

Request a demo Start in the sandbox