auto-reader-ocrlisted
Install: claude install-skill Samqra/auto-reader-ocr
# Auto-Reader OCR API
Self-hosted GPU API, Arabic-first accuracy, PDPL-compliant (images processed
transiently, never stored). Base URL: `https://api.auto-reader.com`.
## Zero-friction start — no human signup needed
If the MCP tools (`ocr_image`, `translate_text`, `ocr_and_translate`,
`get_usage`, `create_api_key`) are available, just call them: the first call
auto-provisions a free trial key and returns it — persist it and pass it back
as `api_key` on later calls.
Otherwise via HTTP:
```bash
curl -X POST https://api.auto-reader.com/signup # {"api_key":"nsk_live_..."} — 25 credits/day
curl -X POST https://api.auto-reader.com/signup -d '{"email":"you@x.com"}' # 250 credits/day
```
## Core calls
```bash
# OCR (1 credit/page; lang auto-detected; mode=document|receipt|manga|scene)
curl -X POST https://api.auto-reader.com/v1/ocr \
-H "Authorization: Bearer $KEY" -F "file=@page.png" -F "lang=auto"
# RAG-ready output (2 credits/page): format=markdown or format=chunks
# markdown -> headings/tables reconstructed; add "Accept: text/markdown" for raw body
# chunks -> [{index,type,heading,text,char_count}] title-bounded, tables kept whole
curl -X POST https://api.auto-reader.com/v1/ocr \
-H "Authorization: Bearer $KEY" -F "file=@page.png" -F "format=chunks"
# Whole PDFs: /v1/document (same format/detail params, per-page + top-level output)
# Structured extraction (5 cr/page; preset=invoice|receipt|id|zatca or schema=<JSON>)
# Every field returns {value, con