ocr image

Extract text from an image with GPU OCR. Best-in-class Arabic (plus Persian/Urdu) accuracy, manga-aware vertical Japanese, and strong English, French, Spanish,

liveverifiedunknown~5s typical

mcpprobedocr

$0.0020

per call

plus $0.001 platform fee per run

Open MCP server; Cracked bills routing only

Provider Upstream docs

$Set up https://cracked-ai-pearl.vercel.app/SKILL.md, then use Cracked to run mcp-com-auto-reader-ocr /ocr_image for me.

Input

Field	Type	Description
image_base64*	string	The image encoded as base64 (a data: URI prefix is accepted and stripped).
lang	string	Language/script hint. Default "auto" detects it. Codes: ar, fa, ur, en, fr, es, de, ja, zh default "auto"
mode	string (document \| receipt \| manga \| scene)	Content hint that tunes detection and prompts. Default "document". default "document"
quality	string (standard \| high)	"standard" (default) lets a confidence gate decide whether the vision model re-reads the p default "standard"
api_key	string	Optional Auto-Reader OCR key (nsk_live_...). If omitted, a free trial key is auto-provisio

Call it

curl

curl https://cracked-ai-pearl.vercel.app/v1/run \
  -H "Authorization: Bearer ck_live_..." -H "content-type: application/json" \
  -d '{"provider":"mcp-com-auto-reader-ocr","endpoint":"/ocr_image","input":{"image_base64":"...","lang":"auto","mode":"document","quality":"standard"}}'

cli

npx cracked-ai run -p mcp-com-auto-reader-ocr -e /ocr_image -i '{"image_base64":"...","lang":"auto","mode":"document","quality":"standard"}'

mcp

run_tool({ provider: "mcp-com-auto-reader-ocr", endpoint: "/ocr_image", input: {"image_base64":"...","lang":"auto","mode":"document","quality":"standard"} })

Try it

Runs against your signed-in workspace balance. Sign in if you have not.