Firecrawl MCP - Developer Guide
This guide explains what Firecrawl MCP is best at, how its toolset is shaped, and when it is the right choice for search, scraping, or broader web extraction workflows.
For the simpler usage view, see the Firecrawl MCP User Guide.
This guide is based on the official firecrawl/firecrawl-mcp-server repository and docs, reviewed on June 26, 2026.
1. What it is​
Firecrawl MCP is a web-focused MCP server for:
- search,
- scrape,
- map,
- crawl,
- extract,
- interactive web actions,
- async research jobs.
It is much broader than a simple page fetcher.
2. Best fit​
Use Firecrawl when the problem is getting structured information out of the live web.
| Task | Why it fits |
|---|---|
| Scrape a known page | Strong single-page extraction |
| Search then extract | Built-in search plus scrape flow |
| Discover site URLs | map is designed for that |
| Run broader research | Async agent mode is built for multi-source work |
3. Setup options​
The upstream docs show multiple paths:
- hosted MCP endpoint,
- hosted keyless free tier,
- local
npxserver, - stdio or HTTP streamable local mode.
Hosted MCP endpoint:
https://mcp.firecrawl.dev/v2/mcp
Local run:
env FIRECRAWL_API_KEY=fc-YOUR_API_KEY npx -y firecrawl-mcp
4. What makes it useful​
Firecrawl is useful because it separates different web tasks cleanly:
- use
searchwhen you do not know the page yet, - use
scrapewhen you know the URL, - use
batch_scrapefor multiple URLs, - use
mapfor site discovery, - use
crawlfor multi-page coverage, - use
extractfor schema-shaped output, - use
agentfor async web research.
That tool separation makes agent behavior more predictable than generic browsing alone.
5. Strengths and limits​
Strengths
- broad web extraction surface,
- strong structured extraction options,
- good docs-capture and research workflows,
- hosted remote option lowers setup friction.
Limits
- large crawl outputs can explode context size,
- rate limits and credits matter,
- not every task needs an autonomous web agent.
6. When to choose it​
Choose Firecrawl when:
- web content is the source of truth,
- you need scraping or structured extraction,
- you want more than a simple search result list.