Firecrawl MCP Server
The web context API for AI agents: search, scrape, crawl, parse documents, and interact with dynamic web pages via hosted or local MCP endpoints.
About this Server
The official Firecrawl Model Context Protocol (MCP) server provides AI agents with an end-to-end web context engine. It allows agents to search the open web, deep developer indexes (GitHub issues, PRs, docs), US government databases, and scientific literature (PubMed, arXiv). Agents can scrape any URL into clean Markdown, structured JSON schemas, or brand design specs; crawl entire domains with depth boundaries; map full site URL graphs; parse local or uploaded PDFs/documents; and interact with dynamic JavaScript applications (clicking, typing, form filling) in real cloud browsers. It also supports autonomous multi-source research agents, recurring page monitoring with automated change diffs, and transparent credit tracking.
Direct low-latency RPC protocol communication with structured JSON schemas.
Restricted capability bounds to prevent accidental modifications or leakage.
Security & Best Practices
When to Use (and When Not To)
- •Extracting clean Markdown or JSON from websites and documentation pages to supply LLM context
- •Executing multi-source autonomous research tasks where target URLs are not known beforehand
- •Searching specialized indices for developer bugs, GitHub PRs, legal statutes, or medical/scientific papers
- •Interacting with dynamic JS-rendered single-page applications requiring clicks or form submissions
- •Setting up recurring monitors with AI change-judging to track competitors, pricing, or docs updates
- •Direct local file editing or project refactoring (use native file tools or Repomix instead)
- •Controlling a local Chrome instance with local session cookies and extensions (use Chrome DevTools MCP instead)
- •High-frequency internal database querying (use PostgreSQL, Neon, or Supabase MCP instead)
Install & Setup
claude mcp add --transport sse firecrawl https://mcp.firecrawl.dev/v2/mcpExposed Tools & Endpoints
13 toolsFunctions your AI agent can invoke via this Model Context Protocol server
Scrapes content from a single URL into clean Markdown, structured JSON adhering to a schema, or brand identity styling.
Searches the web with ranked results, relevant excerpts, optional inline scraped pages, and feedback credit refunds.
Starts an asynchronous crawl job over a website with configurable depth, path filters, and automated status polling.
Fast discovery of all indexed URLs belonging to a target domain without downloading page contents.
Parses local or uploaded PDFs, Word docs, spreadsheets, and HTML files into Markdown or structured JSON.
Interacts with dynamic pages in a live browser (clicks, typing, form filling, navigation) using natural language or code.
Deploys an autonomous web research agent that searches, reads multiple sites, and returns assembled structured data.
Polls the status of an active autonomous research agent job and retrieves structured results upon completion.
Searches an index of GitHub issues, merged PRs, repository READMEs, and code docs specifically tailored for coding agents.
Searches US federal, state, and court primary legal documents, statutes, regulations, and judicial opinions.
Searches scientific paper abstracts and full text across PubMed, bioRxiv, medRxiv, arXiv, and academic repositories.
Creates a scheduled monitor to watch pages for meaningful content changes with automated diff alerts and webhooks.
Checks remaining and historical Firecrawl API credit consumption, broken down by billing period, key, or endpoint.