Quick Start: Document Extraction API

Quick start: your first extraction in 10 minutes.

In this guide, you generate a token, send a document to the Content Extraction endpoint and read the result: the document type, the extracted fields and the confidence score of each field.

Choose your path

I want to...	Path
See the result without writing code	Upload a document in the Demo Portal
Integrate through the API	Follow the steps below

Before you start

POC client key. It has limited usage and is meant for testing. Already a customer? Request one at comercial@most.com.br.

Calls from your backend. The API must be called from your server, never directly from an app or browser. Any communication with apps and devices goes through your backend before reaching mostQI.

A test document. A photo or PDF of a CNH, RG or another document from the catalog, with up to 5 pages.

Not a customer yet?

Store the client key in an environment variable:

How it works

The diagram shows the calls you will make in the steps below. Everything happens from your backend.

Step 1 – Generate the token

INFO

You must call the authentication endpoint before any request to the APIs.

Authentication uses the JWT standard, with single-use, temporary tokens generated by the Authenticate (Get JWT) endpoint from your client key. The token is valid for 10 minutes and can be used only once: generate a new one for each call.

Request: POST https://mostqiapi.com/user/authenticate

{ "token": "<your client key>" }

Expected response:

{ "token": "eyJhbGciOi..." }

Step 2 – Send the document

Send the file encoded in base64 in the fileBase64 field. If the file is available at a public URL, use fileUrl instead.

cURL

Python

JavaScript

On Linux. On macOS, replace base64 -w0 document.jpg with base64 -i document.jpg.

Step 3 – Read the result

The response contains a result list with one item per document found in the file. A single file can contain several documents, such as the front and back of an ID.

Short response example, with illustrative values:

{
  "result": [
    {
      "type": "documento-pessoal",
      "stdType": "personal-document",
      "score": 0.98,
      "pageNumber": 1,
      "tags": ["country=bra", "subtype=rg", "region=verso"],
      "fields": [
        { "name": "nome", "stdName": null, "value": "MARIA DA SILVA", "score": 0.99 },
        { "name": "cpf", "stdName": null, "value": "00000000000", "score": 0.99 },
        { "name": "data_nascimento", "stdName": null, "value": "1986-05-16", "score": 1 }
      ],
      "tables": [],
      "crops": []
    }
  ],
  "requestId": "6ERWpivDfUmhEoqnXeDHi",
  "elapsedMilliseconds": 1089,
  "status": { "code": "200", "message": "Ok", "errors": null }
}

Field	What it is	How to use it
`result`	List with one item per document found in the file	Loop through all items; a file may contain front and back
`type` / `stdType`	Document type, in the original language and in standardized form	Check that it is the document you expected
`tags`	Document characteristics (country, subtype, side)	Use them to separate front and back or filter subtypes
`score` (document)	Overall rating, from 0 to 1, of the image and reading quality	Documents with a low score are candidates for a new capture
`fields[].name`	Name of the extracted field	Always locate fields by name
`fields[].value`	Value read	Store it in your system if the field score meets your threshold
`fields[].score`	Confidence of the field reading, from 0 to 1	Decide whether it flows automatically or goes to review
`requestId`	Request identifier	Share it with support if you have questions

WARNING

Field order is not guaranteed. Always locate a field by its name, never by its position.

Step 4 – Decide based on the score

Each field has a score from 0 to 1, and the document also has an overall score. The higher the score, the more reliable the reading. Use the score to decide the path of each document:

Score at or above your threshold: the data flows automatically into your system.

Score below the threshold: the document goes to human review, by your team or in mostValid.

In the POC, treat fields scoring below 0.5 as candidates for human review. Test with real documents from your operation and track how many items fall below that value. 0.5 is a starting point, not an accuracy guarantee: adjust the threshold to the risk of each field. A higher threshold reduces errors and increases review; a lower one does the opposite.

If the review volume is significant, or if your operation already requires a controlled process, plan to use mostValid. With mostValid, you send documents through its own endpoints, and review tasks are opened automatically according to the rules defined for your operation.

Learn more in Confidence score.

If something goes wrong

Code	HTTP	What it means	What to do
R117	401	Invalid, expired or already used token	Generate a new token for each request
R101	400	Incorrect parameter	Check the field names and send `fileBase64` or `fileUrl`
R103 or R115	400	Invalid base64	Encode the entire file in base64 (RFC 4648)
R113 or R130	400	Unsupported file format	Use a format from the list of supported extensions
E400	400	File too large	Up to 50 MB for PDF and 20 MB for images
R120	408	Processing timed out	For PDFs with more than 5 pages, use the asynchronous endpoint
R110	429	Request limit reached	Wait a few seconds and resend
S101 or S103	400	Account without the service or without credits	Contact comercial@most.com.br

The full list is in Error Messages.

Before going to production

Replace the POC client key with your production client key.

Use the production domain with a whitelist and register your backend IPs (see the table below).

For PDFs with more than 5 pages, use the asynchronous Content Extraction endpoint with a webhook.

Files: up to 50 MB for PDF and 20 MB for images.

Domains

Domain	Use	Access
`production-mostqiapi.com`	Production	Only for IPs registered in the mostQI whitelist
`mostqiapi.com`	Staging and POC	From any IP

Both domains point to the same production environment. Do not use mostqiapi.com in production: it may be interrupted by security triggers.

CAUTION

All requests must use the HTTPS protocol.

Security and responsible use

DANGER

Tests run directly in this documentation must use only the POC client key, which is meant for demonstration and has limited usage. The production client key must be used exclusively in the direct integration from your backend.

Client key security: the customer is responsible for keeping the client key safe and confidential, with appropriate measures against unauthorized access. The customer is responsible for the costs of all queries made with their client key.

Responsible use: use of the API and client keys must follow the terms of the service agreements and your organization's security policies.

Next steps

I want to...	Where to go
See all extraction parameters and options	Content Extraction
Process long files or batches	Asynchronous Content Extraction
Only identify the document type, without extracting	Content Classification
Validate the data against official sources	Enrichment
Confirm identity with a selfie and liveness check	Facematch and Liveness Streaming
Manually review low-score cases	mostValid
Know which documents and fields are read	Support Material
Understand limits, formats and billing	Characteristics

Quick start

Choose your path#

Before you start#

How it works#

Step 1 – Generate the token#

Step 2 – Send the document#

Step 3 – Read the result#

Step 4 – Decide based on the score#

If something goes wrong#

Before going to production#

Domains#

Security and responsible use#

Next steps#

Choose your path

Before you start

How it works

Step 1 – Generate the token

Step 2 – Send the document

Step 3 – Read the result

Step 4 – Decide based on the score

If something goes wrong

Before going to production

Domains

Security and responsible use

Next steps