Can this convert Swagger UI pages that render via JavaScript?

Yes — Swagger UI is fully client-rendered, so we route those URLs through our headless-browser path automatically. The page is rendered to its final state, all collapsed endpoint cards are programmatically expanded, then we convert. See JavaScript SPA to Markdown for the underlying mechanism.

Are request/response schemas preserved as tables?

Parameters and headers become GFM tables (name, location, type, required, description). Request and response body schemas become fenced JSON code blocks with syntax highlighting. The split is intentional: tables for things you scan, code blocks for things you copy.

What if I have the OpenAPI spec directly — should I still use this?

No — if you have the raw openapi.json or openapi.yaml, use a dedicated OpenAPI-to-Markdown generator (widdershins, openapi-markdown) for higher fidelity. This converter is for the rendered HTML pages, useful when you can't access the spec but can access the docs site.

Are code examples in multiple languages preserved?

Yes — Swagger UI, Redoc, and Stoplight all surface code examples in tabs (curl, Python, Node, Go). We emit each tab as a separate fenced code block under a sub-heading (### curl, ### Python) so the language hint is visible and copy-paste works in any Markdown viewer.

Does this work on Stripe, Twilio, or GitHub-style API references?

Yes — those three are bespoke (not Swagger UI), but they all follow the same conventions: endpoint cards with method+path headings, parameter tables, code samples in tabs. The extractor recognises the major patterns. For private docs behind auth, the MDisBetter web tool fetches anonymously — you'll need a self-hosted script using requests with your session cookie and Trafilatura/html2text for the conversion step.

API Docs to Markdown — Swagger, OpenAPI Pages

Why generic scrapers fail on API docs

Two reasons. First, most modern API docs (Swagger UI, Redoc, Stoplight, Mintlify, ReadMe.io) render endpoint detail by JavaScript from an OpenAPI/Swagger spec — a static fetch sees a near-empty shell. Second, even when content is in the HTML, it's organised by interactive cards that hide most data behind "Try it" buttons or collapsed accordions. Markdown has no concept of "collapsed", so a faithful conversion needs to expand everything first.

Endpoint extraction

For each endpoint, we emit: the HTTP method and path as a heading (## GET /users/{id}), a short description, a parameters GFM table (name, in, type, required, description), the request body schema as a code block, response schemas grouped by status code as code blocks, and any code examples in their original languages as fenced blocks. The result reads like a hand-written reference and is searchable with grep.

API Documentation to Markdown — Developer Docs Offline

Why generic scrapers fail on API docs

Endpoint extraction

Before / After

Frequently asked questions

Stop feeding garbage
to your AI

Tools

Stop feeding garbageto your AI

Tools

Stop feeding garbage
to your AI