Firecrawl CLI
Agent skill published by Firecrawl: `npx firecrawl-cli init` installs the CLI and a skill file into Claude Code, Cursor, Codex, Windsurf, OpenCode, OpenClaw, OpenHands, Hermes Agent. Main commands: scrape (single page), crawl (entire site), search (web search), map (list a site's URLs), batch-scrape (URL list as one job), agent (AI-driven extraction and research), interact (actions in a live browser), monitor (change detection).
Strengths
- Multi-agent install in one command: `init` configures Claude Code, Cursor, Codex, Windsurf and four other agents
- About ten commands (scrape, crawl, search, map, agent, interact, monitor...), covers nearly all web needs of an agent
- Published and maintained by the Firecrawl team (110k+ stars on the parent product), stable API, JS rendering handled server-side
- Auth handled locally (API key stored after browser login), no secrets to copy into agent configs
Limitations
- Built for the Firecrawl cloud, large volumes tip into a paid plan. The engine can also be self-hosted (the `--api-url` option), but without the anti-bot layer or screenshots, which stay cloud-only
- Recent skill (early 2026), less field experience than official Anthropic skills
- Requires restarting the agent after `init` for the skill to be discovered
Best for
- Agents that need to read online documentation or explore sites whose content you don't have locally
- Monitoring workflows: an agent scraping N sites per day and summarizing the updates
- Structured extraction (pricing, specs, listings) from HTML pages without writing CSS selectors