← All guides

Locus product guide · Updated

Firecrawl MCP server: tools, setup, limits and cost

Firecrawl's official MCP server gives an agent scrape, search, map, crawl, parse, agent, interact and research tools. Connect the hosted server at https://mcp.firecrawl.dev/v2/mcp-oauth with a browser sign-in, use https://mcp.firecrawl.dev/v2/mcp with an API key, or use that URL with no key for search, scrape and parse only. Calls spend Firecrawl credits: 1 per page scraped and 2 per 10 search results. The free plan gives 1,000 credits a month.

Setup, step by step

  1. Step 1

    Add the hosted server in Claude Code

    This is the OAuth endpoint from the Firecrawl docs. It needs a Firecrawl account but no API key in your config.

    claude mcp add --transport http firecrawl https://mcp.firecrawl.dev/v2/mcp-oauth
  2. Step 2

    Sign in

    Open Claude Code, run /mcp, pick firecrawl and authenticate in the browser. The tool list then shows the full set your plan allows.

    /mcp
  3. Step 3

    Or add it to Codex

    Codex uses the same URL. Run the login command after adding the server.

    codex mcp add firecrawl --url https://mcp.firecrawl.dev/v2/mcp-oauth
    codex mcp login firecrawl
  4. Step 4

    Use an API key for headless agents

    For a server or CI job with no browser, send your key as a bearer token on the /v2/mcp endpoint. Drop the headers block to try keyless mode.

    {
      "mcpServers": {
        "firecrawl": {
          "type": "http",
          "url": "https://mcp.firecrawl.dev/v2/mcp",
          "headers": { "Authorization": "Bearer <FIRECRAWL_API_KEY>" }
        }
      }
    }
  5. Step 5

    Or run it locally

    The open-source server needs Node.js 22 or newer. The docs pin version 3.33.2.

    env FIRECRAWL_API_KEY=fc-YOUR-API-KEY npx -y [email protected]

What tools does the Firecrawl MCP server include?

The four core tools cover most agent work. Credit use per call comes from the Firecrawl pricing page, checked November 10, 2026.

  • firecrawl_parse turns a PDF, document, spreadsheet or HTML file into text.
  • firecrawl_agent runs multi-source research as an async job, and firecrawl_interact clicks, fills forms and navigates a live page.
  • firecrawl_research_* searches PubMed, bioRxiv, medRxiv and arXiv abstracts. firecrawl_developer_search and firecrawl_gov_search answer coding and legal questions from their own indexes.
  • firecrawl_monitor_* runs recurring checks with diffs and notifications.
  • The old Extract tool is deprecated. Use scrape with JSON output for a known page, or agent when Firecrawl must find the sources.
What tools does the Firecrawl MCP server include?
ToolWhat it doesFirecrawl credits
firecrawl_scrapeReads one URL as markdown, or as JSON that matches a prompt or schema1 per page, plus 4 for JSON output
firecrawl_searchSearches the web when you have a query instead of a URL2 per 10 results
firecrawl_mapLists the URLs on a site before you pick pages1 per call
firecrawl_crawlWalks a site or section and returns each page1 per page crawled

Should you connect with OAuth, an API key or no key?

All hosted modes use the same Streamable HTTP service. The mode decides which tools appear and whose limits apply.

Should you connect with OAuth, an API key or no key?
ModeEndpointTools you getBest for
Hosted OAuthhttps://mcp.firecrawl.dev/v2/mcp-oauthFull set, subject to plan and team policyA person at a desk in Claude Code or Codex
Hosted API keyhttps://mcp.firecrawl.dev/v2/mcp with a bearer headerFull set allowed by your planHeadless agents and CI jobs
Keylesshttps://mcp.firecrawl.dev/v2/mcp with no headerfirecrawl_search, firecrawl_scrape and firecrawl_parseTrying Firecrawl before you sign up
Localnpx -y [email protected] over stdio or HTTPAPI-backed tools; local file parsing needs a self-hosted APIClients that launch a process, or a self-hosted Firecrawl

Which Firecrawl plan does the MCP server use?

The MCP server itself is free. Every tool call draws on the credits of the Firecrawl plan you sign in with. Paid plans are billed monthly or yearly and differ in credits and rate limits. Plans checked November 10, 2026.

  • Free covers about 1,000 scraped pages or 500 searches of 10 results.
  • Rate limits per minute for scrape, map and search: Free 10, Hobby 100, Standard 500. Crawl starts: Free 2, Hobby 20, Standard 100.
  • Concurrent browsers: Free 2, Hobby 5, Standard 25.
Which Firecrawl plan does the MCP server use?
PlanCredits a month
Free1,000
Hobby5,000
Standard100,000
Growth500,000
Scale1,000,000

What are the limits and catches?

Most surprises come from credits, not from the MCP layer. These come from the Firecrawl pricing, rate limit and MCP pages.

  • Keyless mode is capped per IP address per day, by both requests and credits. It is for trials, not production.
  • A page that returns 403 or 404 still costs 1 credit.
  • JSON output costs 5 credits a page in total, five times a plain markdown scrape.
  • Unused credits expire each month, except a one-month rollover on Scale and Enterprise.
  • Pay-as-you-go top-ups are only on paid plans. On Free, you wait for the monthly reset.
  • Some tools can be turned off by your plan or a team admin, so the tool list can differ between two accounts.

How do you call Firecrawl through Locus Pro?

Add the Locus Pro MCP server once and your agent can call Firecrawl with no Firecrawl account or key. Each call is charged the catalog price from one prepaid Locus Pro balance, not from a Firecrawl plan. Searching the catalog and reading an endpoint contract are free, and a call can carry max_credits as a price ceiling.

These Firecrawl endpoints were in the Locus catalog on November 10, 2026.

  • MCP URL: https://api.paywithlocus.com/api/credits/mcp
  • Or tell your agent: Install Locus Pro by following https://paywithlocus.com/SKILL.md
How do you call Firecrawl through Locus Pro?
Locus endpointUseHow it is priced
firecrawl/searchWeb searchFixed price per search
firecrawl/mapList the URLs on a siteFixed price per call
firecrawl/scrapeOne page as markdownVaries by request
firecrawl/scrape-v2One page, including JSON extractionVaries by request
firecrawl/crawl and firecrawl/batch-scrapeMany pages in one jobBy the number of pages
firecrawl/extractOlder structured extractionVaries by request

Should you use Firecrawl's MCP server or Locus Pro?

Use Firecrawl's own server if you need monitor, interact, agent, research or government index tools, or if your team already pays for a plan. Use keyless mode to try it, and the free plan for light use. At high, steady volume, pages can cost less with your own Firecrawl account and API key.

Use Locus Pro when volume is low or bursty, when you do not want a monthly plan, or when the same agent also calls other search, data or model APIs. One balance and one connection replace a key per vendor.

Frequently asked questions

Is the Firecrawl MCP server free?

The server is free. Keyless mode allows search, scrape and parse with a daily cap per IP address. Signed-in use spends your Firecrawl credits, and the free plan gives 1,000 credits a month.

Why is a Firecrawl tool missing from my tool list?

Keyless connections only show search, scrape and parse. Signed-in connections can still lose tools to plan limits or a team policy. A local server only lists tools its target API supports.

Can the Firecrawl MCP server read a PDF on my computer?

The hosted server uses a signed upload step for files. A local server can read a file path directly only when it points at a self-hosted Firecrawl API, not the cloud API.

What replaced the Firecrawl extract tool?

Firecrawl deprecated it. Use scrape with JSON output when you know the page, or the agent tool when Firecrawl has to find the sources first.

Implementation references

Use these first-party references for current request contracts and account requirements. Tool availability, prices, and negotiated terms can change.

Sources

Third-party details were checked on November 10, 2026. Vendors change pricing and features often, so confirm on their site before you decide.

Related guides