xTrakt API Reference

The xTrakt API exposes the platform's document intelligence engine (classification, extraction, zero-shot field discovery, cropping and splitting) through a small, focused set of HTTP endpoints. It is the same engine that powers the playground inside the xTrakt platform — wrapped behind API-key authentication so you can call it from your own systems.

Base URL

The development environment is publicly available at:

https://api-dev.xtrakt.ai

All endpoints documented in this section live under this host. Every request is HTTPS-only.

Authentication

Every endpoint requires an API key issued from the xTrakt platform. The key is sent on the standard Authorization header as a Bearer token:

Authorization: Bearer xtkt_live_xxxxxxxxxxxxxxxxxxxxxxxxxxxx

• The key always starts with the prefix xtkt_live_.
• The platform stores only a hash of the key — the plaintext value is shown to you once, at creation time. Store it in a secrets manager.
• A key can be revoked at any time from the platform. Revoked keys return 401 Unauthorized.
• The key is bound to your tenant; no extra x-tenant header is needed.

See Getting Started for the step-by-step flow to generate a key.

Available Endpoints

All endpoints are POST and accept multipart/form-data request bodies.

Endpoint	Purpose
POST /document/classifier	Classifies one or more documents into the document types registered for your tenant.
POST /document/extraction	Extracts the structured fields from a single document.
POST /document/zero-shot	Discovers candidate fields from a document with no prior template.
POST /document/crop	Returns the cropped/normalized version of a document image.
POST /document/split	Splits a long PDF into the logical sub-documents it contains.

Limits & file rules

• Accepted file types: .pdf, .png, .jpg, .jpeg, .tiff, .tif, .bmp, .doc, .docx.
• Maximum file size: 10 MB per file on every endpoint except /document/split, which accepts large PDFs.
• Rate limit: 50 requests per minute, per API key (HTTP 429 Too Many Requests when exceeded).

Response envelope

Responses are always wrapped in the standard xTrakt service envelope:

{
  "data": { /* endpoint-specific payload */ },
  "meta": {
    "httpStatusCode": 200,
    "messages": []
  },
  "hasErrors": false
}

When hasErrors is true, meta.messages contains human-readable error descriptions and data is null.