Skip to content
MCP Shelf

Pod to App: Claude Code

How to install Firecrawl MCP in Claude Code

By Sophie Filer · Updated 11 Nov 2026 · Steps checked against Firecrawl's own docs on 11 Nov 2026

The short answer

Run the command below, then type /mcp in Claude Code and sign in to Firecrawl in your browser. That connects Firecrawl's hosted server with your account's full tool set, and there is no API key to copy around.

Terminal
claude mcp add --transport http firecrawl https://mcp.firecrawl.dev/v2/mcp-oauth
Sign-in
Browser, key, or none
Runs on
Firecrawl's hosted server
No account
Search, scrape, parse only
Price
Free tier, paid plans

Pick how to connect

ModeCommandYou get
Sign in (recommended)claude mcp add --transport http firecrawl https://mcp.firecrawl.dev/v2/mcp-oauth, then /mcpEvery tool on your plan; each connection gets its own key you never see
Keyless trialclaude mcp add --transport http firecrawl https://mcp.firecrawl.dev/v2/mcpSearch, scrape and parse only, rate limited, no account
API keySame keyless URL plus -H "Authorization: Bearer YOUR_FIRECRAWL_API_KEY"Full tools, for scripts and machines with no browser

Install it step by step

  1. Step 1: In your terminal, run the command at the top of this page. Add --scope user after add to have Firecrawl in every project.
  2. Step 2: Start Claude Code and type /mcp. Select firecrawl and choose to authenticate.
  3. Step 3: Sign in to Firecrawl in the browser and approve a team.
  4. Step 4: Back in /mcp, firecrawl should be connected. Try: "Search the web for the latest Firecrawl release notes and summarise the sources."

You can review or revoke the connection later under Connected apps in your Firecrawl settings.

Which Firecrawl tool Claude Code should use

You wantTool
One known page as clean Markdownfirecrawl_scrape
A list of URLs on a site, without the contentfirecrawl_map
Many pages under one sitefirecrawl_crawl, with a limit and include or exclude paths
Results for a query rather than a URLfirecrawl_search
A page that needs a click or login step firstfirecrawl_interact
Credits left this monthfirecrawl_credit_usage

Run it locally instead

The open-source server also runs on your machine through npx, which suits self-hosted Firecrawl or setups that must not call a hosted MCP endpoint:

Terminal
claude mcp add firecrawl -e FIRECRAWL_API_KEY=fc-YOUR_API_KEY -- npx -y firecrawl-mcp

For a self-hosted Firecrawl, add -e FIRECRAWL_API_URL=https://your-firecrawl-host as well.

If it does not work

What you seeWhat to do
Only three Firecrawl toolsYou are on the keyless URL. Switch to the sign-in URL or add an API key header.
A browser page shows an error when you open the URLDo not open it yourself. Run /mcp in Claude Code and authenticate from there.
Two firecrawl entries after reconnectingRemove the extra one with claude mcp remove firecrawl; reconnect through /mcp rather than adding again.
Rate limit messagesKeyless use is capped. Sign in or use a key for higher limits.

MCP Shelf take

Is Firecrawl worth adding to Claude Code?

Claude Code can already fetch a single web page, so Firecrawl earns its place on the harder jobs: whole-site crawls, sites full of scripts, search plus scrape in one step, and pages that need a click first. Results come back as clean Markdown, which saves a lot of context compared with raw HTML. Start keyless to see if it fits your work, then sign in through /mcp; it is the neatest setup, with no key in any file. Watch credits on big crawls and always set a page limit.

Sophie Filer

Questions

Can I use Firecrawl MCP in Claude Code without an API key?

Yes. The keyless URL https://mcp.firecrawl.dev/v2/mcp gives search, scrape and parse within daily limits. Signing in through /mcp unlocks the rest without handling a key yourself.

What is the difference between the mcp and mcp-oauth URLs?

/v2/mcp-oauth signs you in through the browser. /v2/mcp is keyless on its own, or uses an API key when you add an Authorization header.

Where is my Firecrawl API key stored in Claude Code?

With sign-in, nowhere you can see: each connection gets its own key. With the header option, it is saved in your Claude Code config, so prefer --scope user over a project .mcp.json you might commit.

How do I stop Claude Code using Firecrawl?

Run claude mcp remove firecrawl, and revoke the connection under Connected apps in your Firecrawl settings.

Sources: the steps on this page are written in our own words from Firecrawl's own install docs, the README, the README, the README. If they disagree with this page, the project's docs win; please tell us from the server page.