Introduction

DocsRouter is an early-access OCR API for extracting text and tables from images and documents. It exposes OpenRouter's currently priced image-input models behind one authenticated endpoint and one response shape.

What is available today

• Native OCR endpoint: Send a URL or base64-encoded document to POST /v1/ocr.
• Live image-model catalog: Choose any currently listed model that declares image input, text output, and valid pricing.
• Transparent routing policies: Choose quality, accuracy, speed, cost, or balanced. Each policy maps to a fixed model; DocsRouter does not inspect a result and retry with another model.
• Multi-page processing: Supported PDFs and converted office documents return a pages array plus combined text.
• Usage details: Responses report the selected model, token usage, pages processed, processing time, and cost breakdown.

DocsRouter does not currently provide a result cache, confidence-based fallback, automatic provider waterfall, or published accuracy and latency guarantees.

Available models

Read GET /v1/models at runtime for the current catalog. The deterministic routing policies continue to use a small recommended subset; explicit selection can use any returned model ID. See Models for the mappings and current pricing metadata.

Pricing

Vision-model requests are billed at the provider cost plus a 5% DocsRouter platform fee. The API returns the provider cost, platform fee, and total cost for each request.

Mistral OCR's direct provider rate for its dedicated PDF parser is $2 per 1,000 pages before the DocsRouter platform fee. Per-request minimum charges and rounding can affect small jobs.

First request

curl -X POST https://api.docsrouter.com/v1/ocr \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "url": "https://example.com/invoice.png",
    "strategy": "balanced",
    "options": {"output_format": "markdown"}
  }'