TigerOCR

TigerOCR turns documents into Markdown. It is a hosted OCR service around baidu/Unlimited-OCR, a 3B document-to-Markdown vision model: hand it an image or a PDF and get back clean Markdown with tables, LaTeX formulas, reading order and exported figure crops.

Two ways in

• A remote MCP server for Claude.ai, Claude Code, Gemini CLI, Codex and Cursor. The dashboard renders ready-to-paste snippets for each client.
• A REST API for everything else: upload a file, then poll the job.

Sign in with GitHub to get a dashboard, API keys and a wallet.

Tools your agent gets

• ocr_document runs OCR on a file.
• get_ocr_job fetches a result, windowed by page or character offset so a long document never floods the model's context.
• search_ocr_job finds a passage inside a finished result.
• list_ocr_jobs lists recent jobs.
• usage reports the wallet balance and spend.

Pay for the GPU seconds you use

Every account has a wallet, and a job costs the GPU time it actually used. The gateway reserves an estimate when you submit, settles to the measured GPU seconds when the worker finishes, and refunds failed jobs in full. Credit can be topped up through Stripe Checkout.

The founder cohort, the first six accounts, gets $5 of credit every month.