Firecrawl Keyless Search

Problem

Broad searches often return fragmented summaries without enough context to find reliable sources or extract usable text. Firecrawl Keyless Search combines web search and single-page scraping in one Python command flow: query web, image, and news sources with --sources, then convert key URLs into Markdown, HTML, links, or screenshots with firecrawl_scrape.py. This gives models, note tools, or analysis scripts cleaner structured input.

How It Works

• Search: control result count with --limit, filter domains with --include and --exclude, restrict --categories to github, research, or pdf, and choose text, json, or markdown output.
• Scrape: extract a single page with --formats markdown, html, rawHtml, links, or screenshot; use --only-main to keep the main content and remove navigation or footer noise.
• Complex research: start with a broad scan, then run focused searches against authoritative domains, and finally deep-read the 2-3 most relevant URLs.

Boundaries

Keyless mode is used by default, resolving API keys in this order: --api-key, FIRECRAWL_API_KEY, then keyless. It relies on curl and may fall back to urllib; Windows UTF-8 output is handled automatically. HTTP 403 usually means the IP is blocked in keyless mode, so provide an API key. HTTP 429 means rate limiting; reduce --limit or retry later. Free-tier usage is roughly 2 credits per 10 results, so it is not ideal for high-frequency bulk scraping.