OfficialBrowser & Web

Firecrawl

by Firecrawl

Search, scrape, crawl and interact with web pages from an assistant, with output as Markdown or structured JSON.

Connect Firecrawl

Choose your client

Local

This server runs on your machine. Pick your client for the matching config; placeholders stand in for any secrets.

Important

Some options bill extra: firecrawl_search with scrapeOptions fetches every result's page and bills each one as a scrape.

claude_desktop_config.json
{
  "mcpServers": {
    "firecrawl": {
      "command": "npx",
      "args": [
        "-y",
        "firecrawl-mcp"
      ],
      "env": {
        "FIRECRAWL_API_KEY": "<YOUR_FIRECRAWL_API_KEY>"
      }
    }
  }
}
  • Replace each placeholder with your own value. The registry lists these variables but does not say which are optional.
  • Add this under the existing mcpServers key if the file already has one, then fully restart Claude Desktop.

Format checked Oct 30, 2026 against Claude Desktop documentation. Client settings change, so confirm with the client's own docs.

Connection details

Hosted endpoint
https://mcp.firecrawl.dev/v2/mcp
Package

npm package · stdio
firecrawl-mcp@3.36.0

Overview

The Firecrawl MCP server gives an assistant web access: searching, scraping pages into Markdown or structured JSON, parsing PDFs and documents, crawling or mapping sites, and interacting with pages in a real browser. It can be used through Firecrawl's hosted endpoint or run locally with npx, and the project supports both Firecrawl's cloud and self-hosted deployments. The documentation describes keyless, OAuth and API-key ways to connect depending on the endpoint, and some endpoints expose only a fixed subset of tools.

What it can do

  • Turns any URL into Markdown or structured JSON, and can parse PDFs and other documents.
  • Works with Firecrawl's cloud API or with a self-hosted Firecrawl instance.
  • A search-only hosted endpoint exposes a fixed set of nine tools.

What you need

Sign-in and credentials

On the hosted keyless free tier scrape, search and parse work without an API key (rate-limited); crawl, map and agent need a key, and OAuth or an API key unlocks the full tool set.

Environment variables

Set these before the server starts.

  • FIRECRAWL_API_KEYsecret
Request headers

The hosted endpoint asks for these headers.

  • Authorizationsecret
Runs as

npm package

Sources

MCP Registry
Listed in the MCP Registry as io.github.firecrawl/firecrawl-mcp-server
Last checked
Oct 31, 2026
Record reviewed
Oct 31, 2026

Statements on this page come from

Why is this labelled Official?

Registry name io.github.firecrawl/firecrawl-mcp-server. The registry only lets an organization's owners publish under its io.github.<org> namespace (registry authentication docs, read 2026-10-30). Repository firecrawl/firecrawl-mcp-server: its About description begins "Official Firecrawl MCP server" (read 2026-10-30); the README body does not describe itself as official. Setup is documented at docs.firecrawl.dev.

Listing on HistChat is not an endorsement or a security review. MCP servers can access data or perform actions, so review the source and permissions before connecting.

  • Playwright

    MicrosoftOfficial

    Give an assistant control of a real browser to navigate pages, click, type and read page structure.

    Category
    Browser & Web
    Connection
    Local
    GitHub stars
    39k
  • Perplexity

    PerplexityOfficial

    Give an assistant Perplexity's web search, grounded answers and deeper research through a hosted or local server.

    Category
    Search & Research
    Connection
    Remote · Local
    GitHub stars
    2.5k
  • Brave Search

    BraveOfficial

    Add web, news, image, video and local search to an assistant using the Brave Search API.

    Category
    Search & Research
    Connection
    Local
    GitHub stars
    1.5k