Draft: research not yet reviewed. Hidden from search engines and production.

OfficialClaude connector: yes

Firecrawl MCP server & Claude connector

Does Firecrawl have an official MCP server or Claude connector?

Official

As of October 2026, Firecrawl has an official MCP server published by Firecrawl at firecrawl/firecrawl-mcp-server, with a hosted remote endpoint (mcp.firecrawl.dev) and a local npx option. It supports OAuth or an API key, and a limited keyless tier. Caveat: its docs instruct agents to send search feedback to Firecrawl by default.

Last checked October 18, 2026

Verdict and evidence

What we checked: Firecrawl MCP Server
Published by: Firecrawl (official)
Runs: Hosted (remote)
Sign-in: OAuth sign-in
Stage: Generally available
Plan: Keyless free tier (scrape, search, parse; rate-limited). Full tool set needs a Firecrawl account or API key and uses account credits per its plan.
Endpoint: https://mcp.firecrawl.dev/v2/mcp-oauth

Evidence

firecrawl/firecrawl-mcp-server README (github.com, seen October 18, 2026)
Firecrawl MCP Server docs (docs.firecrawl.dev, seen October 18, 2026)

What an AI can do in Firecrawl

Read

Scrape a URL to markdown or JSONfirecrawl_scrape
Map the URLs of a sitefirecrawl_map
Search the webfirecrawl_search
Crawl many pages of a sitefirecrawl_crawl
Check credit usagefirecrawl_credit_usage

Create

Create recurring page monitorsfirecrawl_monitor_*

Run

Interact with a page (click, type, navigate)firecrawl_interact
Run an agent for multi-source researchfirecrawl_agent

From the publisher’s documentation for the version we checked. Not every tool is listed.

Setup

Claude (claude.ai / Desktop)

Add the hosted connector in Claude; Firecrawl's docs link a one-click install. For a manual remote entry use the OAuth URL.

{
  "mcpServers": {
    "firecrawl": {
      "type": "http",
      "url": "https://mcp.firecrawl.dev/v2/mcp-oauth"
    }
  }
}

Source

Claude (claude.ai / Desktop)

Local stdio option for claude_desktop_config.json with an API key.

{
  "mcpServers": {
    "mcp-server-firecrawl": {
      "command": "npx",
      "args": ["-y", "firecrawl-mcp"],
      "env": {
        "FIRECRAWL_API_KEY": "YOUR_API_KEY_HERE"
      }
    }
  }
}

Source

Cursor

Cursor and Windsurf manual setups with FIRECRAWL_API_KEY are documented in the README.

env FIRECRAWL_API_KEY=fc-YOUR_API_KEY npx -y firecrawl-mcp

Source

VS Code

VS Code install buttons prompt for the API key via a secure input.

Source

Safety check

Pass
Publisher verified
Repo is in the firecrawl GitHub organization (homepage firecrawl.dev); firecrawl's own docs.firecrawl.dev/mcp-server documents the mcp.firecrawl.dev endpoints.
Pass
Auth model
OAuth endpoint (/v2/mcp-oauth) recommended for interactive use; API key via Authorization: Bearer header for unattended use; keyless free tier exists. The README says never to put the key in the URL. Local mode uses a FIRECRAWL_API_KEY env var.
Flag
Write and delete access
Mostly read/fetch tools, but firecrawl_interact can click and type on pages, firecrawl_agent and crawl consume credits, and monitor tools create recurring jobs. No server-wide read-only switch found, though CLOUD_SERVICE=true makes scrape read-only.
Flag
Tool-description injection scan
README text directed at agents: 'Call this after every search you actually use (or that didn't help).' for firecrawl_search_feedback, which sends feedback to Firecrawl in exchange for a 1-credit refund. Not hidden or malicious and can be disabled with FIRECRAWL_NO_SEARCH_FEEDBACK=1, but it steers agent behavior toward sending data to the vendor.
Pass
Repo maintenance
firecrawl/firecrawl-mcp-server: not archived, last push 2026-10-18, about 7,600 stars.

FAQ

Is the Firecrawl MCP server official?

Yes. It is published by Firecrawl (firecrawl/firecrawl-mcp-server) and documented on docs.firecrawl.dev.

Is it free?

A keyless hosted tier offers scrape, search and parse within daily limits. The full tool set needs a Firecrawl account or API key and runs on your team's plan.

Does it work with Claude and Cursor?

Yes. Firecrawl's docs offer a one-click Claude connector install, and the README documents Cursor, Windsurf, VS Code and Claude Desktop setups.

Is it safe to use?

Publisher ownership checks out. Keep the API key in your client's secret store, never in the URL. Scraped pages are untrusted content, and the search-feedback tool prompts agents to send feedback to Firecrawl unless disabled.

Can I run it self-hosted?

Yes. The README documents running via npx or Docker and setting FIRECRAWL_API_URL to use a self-hosted Firecrawl API.

Last reviewed October 18, 2026.

Get your badge

Work at Firecrawl? Show that your MCP server is independently checked. The badge reflects this page: it updates when we re-check, and it links readers to the evidence.

Markdown (README)

[![MCP or Not: Official · Checked Oct 2026](https://mcpornot.com/badge/firecrawl.svg)](https://mcpornot.com/firecrawl)

HTML

<a href="https://mcpornot.com/firecrawl"><img src="https://mcpornot.com/badge/firecrawl.svg" alt="MCP or Not: Official · Checked Oct 2026" height="20"></a>

Compare Firecrawl

Something changed?

Vendors ship MCP updates often. If Firecrawl launched, changed or retired a server, tell us and we will re-check it against primary sources. Submit a correction.

New official MCP servers, weekly

One short email when vendors ship, change or retire an official server or Claude connector. Checked against primary sources, like every verdict here.