PDFtoMarkdown.ai

Drop a file into the converter and adjust three things: OCR, AI assistance, and image extraction. The first ten pages are free with no signup, which is enough to check whether your particular documents survive the trip.

For anything ongoing you connect through the REST API or the MCP server, the latter being the route that lets Claude, Cursor, Codex or a local agent call the converter as a tool without you brokering the file. Output is Markdown with the document structure preserved, which is the part that matters for retrieval: a table that stays a table can be queried, while a table rendered as loose text cannot.

Billing runs on pages. Credits never expire, and one credit covers one standard page, with OCR and AI options costing more.

• •MCP server - an agent calls the converter directly as a tool, which removes the manual step of converting a file and pasting the result into a prompt.
• •REST API - the same conversion inside a pipeline, for RAG ingestion or a batch of documents rather than one at a time.
• •Structure preserved, not flattened - headings, tables and extracted images survive as Markdown, which is what makes the output usable for retrieval rather than just readable.
• •OCR for scanned documents - text locked inside an image-only PDF is recovered, at a higher credit cost than a text-native page.
• •Files are not stored - documents are converted and discarded rather than retained, which is the difference between a tool you can point at a client contract and one you cannot.
• •Ten free pages with no signup - enough to test your own worst document before creating an account or paying anything.

Anyone building a RAG pipeline who has discovered that PDF ingestion is where retrieval quality is actually decided. Developers running coding agents that keep being handed specifications, standards documents or manuals as PDFs. Researchers and students converting academic papers into something a model can quote accurately rather than approximate.

The sharpest fit is the agent case, because that is where the cost argument lands hardest. An agent that re-OCRs the same document on every run pays for it every time; converting once and keeping the Markdown is cheaper and more consistent. If you convert a handful of documents a month, the free pages and a small credit top-up will cover it and there is nothing to subscribe to.