PDF to Markdown, built for AI agents

Convert PDF to Markdown via API. Text-native PDFs take a fast path, scanned PDFs fall back to OCR automatically. Signed URLs up to 500MB, token-aware headers, no LLM in the loop.

Convert a PDF

How it works

01

Upload the PDF

Raw upload or signed URL, no multipart overhead. Files up to 500MB.

02

Markpdf inspects it

Text-native pages route to a fast Cloudflare Edge worker. Scanned or image-heavy pages route to the Full Engine (PyMuPDF / MarkItDown) with OCR — only when the document actually needs it.

03

Get clean Markdown back

Direct text/markdown response, token-aware headers so your agent can weigh cost vs. quality before parsing.

FAQ

Does it handle scanned PDFs, not just text-based ones?

Yes. Markpdf checks whether a PDF is text-native first. If it is, conversion is near-instant on the Edge path. If the PDF is scanned or image-based, it automatically falls back to OCR through the Full Engine — no manual flag needed.

What's the max file size?

Up to 500MB via signed-URL streaming, so you're not stuck with multipart upload limits on large PDFs.

Is this built for AI agents specifically?

Yes. Responses come back as direct text/markdown (not wrapped in JSON), and token-aware headers let an agent decide whether to fetch the full document or a cheaper summary path.