Convert PDF API documentation to OpenAPI 3.1
Plenty of APIs still ship their only documentation as a PDF: payment gateways, banks, logistics carriers, ERP vendors, government agencies. The endpoints are real, the parameters are real — but none of it is machine-readable, so every integration starts with someone hand-transcribing tables into YAML.
pdf2api reads the PDF the way an engineer would: it finds endpoint definitions, walks parameter tables (including merged cells and multi-page tables), detects auth schemes, and emits a structured OpenAPI 3.1 document you can lint, diff and feed into your toolchain.
Why hand-writing OpenAPI from a PDF fails
A 50-page API reference typically holds 30–80 endpoints and several hundred parameters. Transcribing that by hand takes an afternoon on a good day — and the result silently drifts from the source: a missed required flag, a wrong type, an enum value typo. Those errors surface days later as integration bugs that are miserable to trace back.
Automated extraction does not make judgment errors of that kind. Where the source document itself is ambiguous, pdf2api keeps the ambiguity visible instead of guessing silently.
What the extraction actually handles
Real-world API PDFs are messier than any demo document. These are the cases pdf2api was specifically built around:
- •Parameter tables with merged cells, nested rows and page breaks in the middle of a table
- •Auth described in prose — API keys, HMAC request signing, OAuth-ish token flows
- •Mixed-language documents (Korean/English) and vendor-specific terminology
- •Request/response examples that disagree with the parameter table (flagged as warnings)
- •Things OpenAPI cannot express — body encryption, signature recipes — surfaced as explicit "implement by hand" warnings instead of being dropped
From upload to a usable spec
Drop the PDF, get a free preview of detected endpoints and auth schemes within about thirty seconds. If the preview looks right, pay once for that document and export the complete OpenAPI 3.1 file as JSON or YAML — plus a Postman collection and Java/PHP/TypeScript SDK drafts generated from the same spec.
The output is a draft by design. You review it, fix what the source document got wrong, and you still come out hours ahead of manual transcription.
Frequently asked questions
Does it work on scanned PDFs?
Text-based PDFs work best. Scanned image-only PDFs depend on the scan quality — try the free preview first; you pay nothing until you export.
Which OpenAPI version is generated?
OpenAPI 3.1, as JSON or YAML. The spec passes standard linters; vendor-specific gaps are annotated under x-pdf2api-warnings.
What happens to my uploaded PDF?
It is kept only until the conversion finishes, then permanently deleted. Unpaid uploads are wiped within 24 hours.
How accurate is the output?
It is an engineered draft, not a guaranteed mirror. Endpoints, parameters and types come straight from the document; anything ambiguous or inexpressible in OpenAPI is flagged as an explicit warning so nothing is silently lost.
How much does it cost?
Preview is free for every document. Exporting costs $5 (≤10 pages), $12 (≤40 pages) or $29 (larger), one-time per document.
Try it on your gnarliest API PDF
Free preview — see detected endpoints and auth before paying anything.
Upload a PDF →