OCR

Active

Not all agents can read text off a scanned page; this extracts it as layout-aware markdown, tables included.

1 tool

Extract text from scanned PDFs and images as layout-aware markdown, including tables. Powered by Mistral OCR. Priced per page.

Documentsocrtext-extractionscanned-pdfmarkdowntables

Tools (1)

▶extract

Extract layout-aware text and tables from a scanned PDF or image.

Usage-based · 1.5 credits per page

Example prompts

“Extract the text from this scanned PDF”
“OCR this photo of a receipt”
“Read the tables out of this scanned report as markdown”
“Get the text from this image”

Parameters

pagesarrayoptional

Optional 0-indexed page numbers to process. Defaults to all pages.

outputstringoptionaldefault: "markdown"

Preferred output style. The response always includes per-page markdown.

input_urlstringrequired

Pre-signed GET URL of the scanned PDF or image.

API Usage

curl -X POST "https://skill.askfaro.com/skills/ocr/run" \
  -H "Authorization: Bearer faro_<your_key>" \
  -H "Content-Type: application/json" \
  -d '{
  "intent": {
    "prompt": "Extract the text from this scanned PDF"
  }
}'

CLI Usage

askfaro describe ocr/extract

Install pip install askfaro-cli, then askfaro auth login.

▶README

OCR

Turn scanned PDFs and photographed documents into accurate, layout-aware markdown (including tables), using Mistral OCR. Faro proxies the call; your file never touches the Faro worker.

Workflow

POST /uploads/presign on faro-api, PUT your file, and pass the get_url as input_url.
The response contains a pages array; each page has markdown text. The number of pages returned is the number of pages billed.

Pricing

1.5 credits per page processed. The response reports the page count, so the charge scales with document length. (The exact response shape is confirmed against the live API once the key is provisioned.)

When to use this vs pdf-tools

Use pdf-tools/extract_text for PDFs that already have a text layer (cheaper, no per-page vendor fee). Use this for scanned or photographed documents, or when you need high-fidelity tables as markdown.

Notes

The default request treats the input as a document URL. Image-only inputs may require the image_url document type, confirmed at finish-line testing once the key is provisioned.