Bluefield
Choosing a scrape-to-markdown API
The best scrape-to-markdown API turns any URL into clean Markdown your model can read directly. Bluefield does this and topped a blind, label-stripped multi-judge quality panel (58.3), returning leaner output than Firecrawl or Jina — and it signs every live result so you can verify the source. Compare candidates on extraction quality, latency, and whether you need provenance.
What is a scrape-to-markdown API?
A scrape-to-markdown API fetches a URL, extracts the main content, strips boilerplate, and returns Markdown — the format LLMs and RAG pipelines consume most cleanly.
- Markdown output
- Clean text optimized for LLM and RAG consumption.
- Boilerplate removal
- Stripping nav, ads, and chrome so only the main content remains.
- Provenance
- A signed, verifiable record attached to the Markdown result.
Bluefield capabilities
| Capability | Detail |
|---|---|
| POST /scrape | Clean Markdown by default; also HTML, links, screenshot, structured JSON |
| Quality | Top of a blind multi-judge panel; leaner output (caveat: single-vendor judges) |
| Signed results | Every live fetch is signed + verifiable (cache hits and PDFs unsigned in v1) |
| JS rendering | Automatic browser escalation for SPA / JS-heavy pages |
| Fast median | 783 ms p50 in our latest run (see /benchmarks for p95) |
Frequently asked questions
- How is Markdown quality measured?
- On open extraction benchmarks vs trafilatura (CleanEval and Dragnet wins, two ties, one loss — we never claim we "beat trafilatura") and a blind multi-judge panel where Bluefield scored highest. The panel uses single-vendor judges and is not yet human-calibrated.
- Does it handle JavaScript-rendered pages?
- Yes — pages that need rendering escalate to the browser tier automatically. That improves coverage but raises latency and cost on those pages.
- Can I prove where the Markdown came from?
- Yes. Every live result ships a signed attestation you verify offline with npx bluefields verify.