Product · Scrape
Any URL to clean Markdown.
Give Webfold a URL or a saved page. You get text a person or a model can read. Structure stays. Clutter goes.
See it work
Try it from wherever you work.
1webfold scrape https://example.com
2webfold scrape saved-page.html --json | jq '{bytes_markdown, blocks_content}'
What you get
What Scrape does.
No browser in the loop
Pages are read directly, so a scrape does not wait for a browser to start or paint. Plain HTML conversion needs no browser process at all.
Links and tables survive
Links are resolved against the page address. A table stays one block with its header row, so a model sees a table, not loose text.
Fallbacks under a budget
If the first fetch comes back empty, Webfold can try a documentation specification or its built-in page model, within an action budget you set.
Chunks and JSON on demand
Split output into retrieval-sized chunks or ask for one JSON object. Progress goes to stderr, so the output pipes cleanly.
How it works
Three steps.
- 1
Fetch
Native HTTP under a deadline and an action budget. Redirects and retries are outside the budget.
- 2
Extract
Scan the HTML into Markdown blocks. Decode embedded page data on thin pages.
- 3
Return
Markdown, JSON or chunks, with links resolved.
Limits up front.
Know the limits
- Plain HTML conversion does not run JavaScript. For pages that only show content after scripts run, connect a browser you control (see Browser).
- The output is untrusted text taken from someone else's page. Sanitise it before showing it anywhere or handing it to a model you do not trust.
- Extraction can miss or keep the wrong blocks. Measured quality is on the Benchmarks pages.
Keep reading.
FAQ
Questions.
Does it run JavaScript?
Plain conversion does not need to. When a page builds itself with scripts, Webfold's own engine runs them and you extract from the result. It passes all 1,928 tasks of Lexbench, the independent agent-browser benchmark. For pages that need canvas, WebGL, web workers or WebAssembly, connect a Chromium you run.
Can I scrape a saved file?
Yes. webfold md saved-page.html converts a file with no network.
What does it cost?
The command-line tool runs on your machine. Hosted pricing is coming soon.
Use Scrape with Webfold.
Run the command-line tool today. Join the waitlist for the hosted API.