Head to head
Firecrawl vs playwright-mcp
Firecrawl and playwright-mcp solve the same job (Agent tools & orchestration), so this is a pick-one decision. If cost is the constraint, playwright-mcp clears it without a card.
Quick read
- Free with no card: playwright-mcp.
| Field | Firecrawlgithub.com Not yet swept | playwright-mcpgithub.com Not yet swept |
|---|---|---|
| What it is | Turns arbitrary sites into LLM-ready markdown or schema-constrained JSON, handling JS rendering, crawling and cleanup. Scrape, crawl, map, batch, search-with-full-content, browser interaction and autonomous agent modes. Parses web-hosted PDFs and DOCX. Self-hostable (AGPL core). ~128k stars. | Drives a real browser and exposes the page to the model as a structured accessibility snapshot rather than a screenshot — so the agent targets elements deterministically without vision or pixel guessing. Adds network mocking, storage state, tracing, video and PDF generation. ~36k stars. |
| Category | Agent tools & orchestration | Agent tools & orchestration |
| Cost tier | mixed | free |
| Pricing | Free 1,000 credits · $16–$599/mo | Free |
| Why builders pick it | The archive-migration tool. Point it at an existing site and get every back-issue article as clean structured JSON ready for Supabase. Also the competitive-research tool — crawl other publications and extract their section taxonomy. Free tier is 1,000 credits; a typical archive fits inside $16/mo. | This is how Claude checks its own work. Have it load the article page, verify the layout at three breakpoints, confirm images lazy-load, and catch the broken nav before you do. The accessibility-tree approach doubles as a free a11y audit on a publication site. |
| Watch out for | not logged | not logged |
| How to wire it up | npx -y firecrawl-cli@latest init --all | claude mcp add playwright npx @playwright/mcp@latest |
| Editor's pick | Yes | Yes |
Other matchups worth running
Agent endpoint: GET /api/public/compare?slugs=firecrawl,playwright-mcp