Evaluation record · mcp-server-firecrawl

Firecrawl MCP Server

v2026.7

Firecrawl

MCPweb-scrapingcrawlingsearchextraction

Strong

About This MCP

Official Firecrawl MCP server (firecrawl-mcp v3.x) providing web scraping, crawling, site mapping, web search, structured extraction, document parsing, a research agent, interactive browser sessions, and page-change monitors. Runs as an npm package over stdio or as the hosted server at https://mcp.firecrawl.dev/{API_KEY}/v2/mcp; a rate-limited keyless free tier (https://mcp.firecrawl.dev/v2/mcp) covers scrape, search, and interact, and hosted transport accepts OAuth bearer tokens.

Last Evaluated: July 9, 2026

Official Website

Trust Vector Analysis

Dimension Breakdown

🚀Performance & Reliability

api reliability

API stability and service maturity analysis

Evidence

Firecrawl Documentation — Mature hosted API backing all MCP tools, with documented status reporting and a widely used production service

highVerified: 2026-07-09

scrape success rate

Scrape success testing across diverse site types

Evidence

Firecrawl MCP Server Repository — Scrape and batch scrape tools handle JavaScript-rendered pages, anti-bot mitigation, and multiple output formats; some heavily protected sites still fail

mediumVerified: 2026-07-09

crawl completeness

Crawl coverage assessment against known site structures

Evidence

Firecrawl Crawl Documentation — Crawl and map tools discover most site pages with configurable depth and limits; very large or dynamic sites may be partially covered

mediumVerified: 2026-07-09

rate limit handling

Rate limiting behavior review from source and docs

Evidence

Firecrawl MCP Server Repository — Built-in automatic retries with exponential backoff for rate-limited requests and credit usage monitoring

mediumVerified: 2026-07-09

error recovery

Error handling and recovery testing

Evidence

Firecrawl MCP Server Repository — Errors surfaced as structured tool results; retry logic covers transient failures, though long-running crawls can require manual restart

mediumVerified: 2026-07-09

🛡️Security

authentication security

Authentication mechanism review

Evidence

Firecrawl MCP Server Repository — Authenticates with a firecrawl.dev API key via environment variable (stdio) or embedded in the hosted endpoint path; hosted HTTP transport also accepts OAuth access tokens (fco_ prefix) via Authorization Bearer header; a keyless rate-limited free tier exposes scrape, search, and interact without credentials

highVerified: 2026-07-09

prompt injection exposure

Threat modeling of untrusted web content returned to the model

Evidence

MCP Security Best Practices — All fetched web content is untrusted; malicious pages can carry indirect prompt injection payloads that the consuming agent must defend against

highVerified: 2026-07-09

ssrf protection

SSRF risk analysis of URL-driven tool inputs

Evidence

Firecrawl MCP Server Repository — Model-supplied URLs are fetched server-side; SSRF-style requests against internal or private endpoints are a documented risk class for fetch/scrape MCP servers

mediumVerified: 2026-07-09

credential handling

API key storage and exposure analysis

Evidence

Firecrawl MCP Setup Documentation — API key stored in client configuration or environment; key grants full account credit usage if leaked, so it must be protected

mediumVerified: 2026-07-09

unauthorized action risk

Capability and blast radius assessment of exposed tools; score lowered from 70 (2026-06-10) because interact browser automation and autonomous agent/monitor tools extend the surface beyond read-only retrieval

Evidence

Firecrawl MCP Server Repository — Core tools remain read-oriented, but v3 adds firecrawl_interact browser automation (click, type, navigate on arbitrary sites, enabling form submissions), an autonomous research agent, and monitor tools that create recurring jobs with webhooks; the agent can also crawl arbitrary sites and consume account credits

mediumVerified: 2026-07-09

🔒Privacy & Compliance

data exposure

Data flow analysis of scrape and crawl pipelines

Evidence

Firecrawl Documentation — Scraped URLs and page content flow through Firecrawl's cloud service and are then returned to the LLM provider context

highVerified: 2026-07-09

sensitive data protection

Privacy controls assessment of returned content

Evidence

Firecrawl MCP Server Repository — No built-in PII or secret filtering on scraped content; anything reachable at a URL can land in the model context

mediumVerified: 2026-07-09

third party data sharing

Data sharing and policy review

Evidence

Firecrawl Privacy Policy — Request URLs and scraped content processed by Firecrawl per its privacy policy; results are additionally shared with the chosen LLM provider

mediumVerified: 2026-07-09

compliance posture

Compliance documentation review

Evidence

Firecrawl Terms of Service — Commercial terms published; users remain responsible for robots.txt, copyright, and lawful-scraping compliance for targets they crawl

lowVerified: 2026-07-09

👁️Trust & Transparency

documentation quality

Documentation completeness review

Evidence

Firecrawl MCP Documentation — Detailed setup guides for major MCP clients, tool-by-tool reference, and configuration options

highVerified: 2026-07-09

operation visibility

Logging and traceability assessment

Evidence

Firecrawl Dashboard — Account dashboard shows request activity and credit usage; MCP tool calls are visible in client logs

mediumVerified: 2026-07-09

open source transparency

Source code and license review

Evidence

Firecrawl MCP Server Repository — MIT-licensed open source server with 6,895 GitHub stars; backend scraping service is partially proprietary

highVerified: 2026-07-09

api coverage clarity

Tool surface documentation review

Evidence

Firecrawl MCP Server Repository — Tool set clearly enumerated in v3: scrape, map, search, crawl (+status), parse, extract, agent (+status), interact browser sessions (+stop), research (papers/repos), monitor (create/list), and feedback tools

highVerified: 2026-07-09

⚙️Operational Excellence

ease of setup

Setup complexity assessment

Evidence

firecrawl-mcp npm package — Single npx command with one API key environment variable, zero-install hosted remote endpoint (mcp.firecrawl.dev), or fully keyless rate-limited free tier for scrape/search/interact

highVerified: 2026-07-09

api performance

Latency characterization across tool types

Evidence

Firecrawl Documentation — Single-page scrapes typically complete in seconds; full crawls of large sites are long-running asynchronous jobs

mediumVerified: 2026-07-09

reliability

Uptime and incident history analysis

Evidence

Firecrawl Status Page — Public status page with historically high availability for the hosted API

mediumVerified: 2026-07-09

feature coverage

Feature completeness assessment

Evidence

Firecrawl MCP Server Repository — Covers the full web data lifecycle: discovery (map, search), retrieval (scrape, crawl), structured extraction, document parsing, autonomous research agent, interactive browser sessions, recurring page monitors, and academic paper/repo research tools

highVerified: 2026-07-09

community adoption

Community activity and adoption analysis

Evidence

GitHub Repository Metrics — 6,895 GitHub stars, active releases (v3.22.3 published 2026-07-08), and listing in major MCP client directories indicate strong adoption

highVerified: 2026-07-09

Strengths

+Full web data toolkit: scrape, crawl, map, search, structured extract, document parsing, research agent, browser sessions, and page monitors
+Handles JavaScript-heavy sites and returns clean LLM-ready markdown
+Automatic retries with exponential backoff and credit usage monitoring
+Simple setup via npx package, hosted remote server, or keyless free tier
+MIT-licensed open source server with strong community adoption (6,895 stars)
+Hosted transport supports OAuth bearer tokens in addition to API keys

Limitations

!All fetched web content is untrusted and can carry indirect prompt injection payloads
!Model-supplied URLs create SSRF-style risk against internal or private endpoints
!Leaked API key allows arbitrary credit consumption on the account
!No built-in PII or secret filtering on scraped content
!Scraped data transits Firecrawl's cloud before reaching the model
!Heavily bot-protected sites can still fail or return partial content
!STATUS 2026-07-09: v3 interact tool adds browser automation (click, type, navigate), so the tool surface is no longer purely read-only — injected instructions could drive form submissions on third-party sites; agent and monitor tools run asynchronously and can consume credits over time

Metadata

license: MIT

supported platforms

0: All platforms with Node.js

1: Hosted remote server

programming languages

0: TypeScript

github repo: https://github.com/firecrawl/firecrawl-mcp-server

github stars: 6895

package name: firecrawl-mcp

package version: 3.22.3

api dependency: Firecrawl API v2 (firecrawl.dev)

authentication: Firecrawl API key (env var or URL path) or OAuth bearer token (hosted); keyless free tier for scrape/search/interact

remote endpoint: https://mcp.firecrawl.dev/{FIRECRAWL_API_KEY}/v2/mcp (keyless: https://mcp.firecrawl.dev/v2/mcp)

maintained by: Firecrawl

transport types

0: stdio

1: streamable-http (hosted)

installation methods

0: npm

1: hosted endpoint

Use Case Ratings

research assistant

Excellent for gathering, crawling, and extracting web sources during research workflows

data analysis

Structured extract and batch scrape turn web pages into analyzable datasets

content creation

Strong for sourcing reference material, competitor pages, and up-to-date facts for content

code generation

Useful for pulling live documentation and examples into coding sessions

customer support

Can crawl knowledge bases and product pages to ground support answers

financial analysis

Good for collecting public filings and market pages, but verify accuracy of scraped figures

legal compliance

Scraping legality and content reliability require careful human review for legal work

education

Helpful for building course material from live web sources with proper attribution

Similar MCPs

MCP Tavily Server

Tavily

MCP Brave Search Server

Anthropic (Archived)

MCP Fetch Server

Anthropic

Apify MCP Server

Apify

Playwright MCP

Microsoft