Firecrawl MCP

When to choose Firecrawl MCP

Use the HTTP API in your application

For request bodies, output formats and response checks, read the Firecrawl API guide.

Connect to Firecrawl MCP

Choose a connection profile below and replace credential placeholders in your own MCP client. The transport describes the MCP connection.

{
  "mcpServers": {
    "firecrawl": {
      "url": "https://mcp.firecrawl.dev/v2/mcp"
    }
  }
}

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

{
  "mcpServers": {
    "firecrawl": {
      "url": "https://mcp.firecrawl.dev/v2/mcp",
      "headers": {
        "Authorization": "Bearer YOUR_FIRECRAWL_API_KEY"
      }
    }
  }
}

Find a page, choose the output, then validate it

A suggested sequence using the documented tools. Check the connected server's tool schema before making a call.

1. Check the connection and tool list

   Save one profile in Cursor, complete any required sign-in, and inspect the discovered tools. A saved JSON file alone does not establish a working connection.

   Sources: Firecrawl keyless and bearer-key configuration, Firecrawl OAuth account connection, Cursor MCP configuration and authentication

2. Find the URL only if needed

   Use Search when the source is unknown. If you already know the page, proceed to Scrape; a search result is a lead to inspect, not evidence that its content answers the task.

   Tools: firecrawl_search

   Sources: Firecrawl current MCP tools and troubleshooting

3. Request the useful output

   Use Scrape for content from the selected URL. For fixed fields, choose its JSON format and supply the client schema or prompt. Check the returned target content before accepting it; extraction can add credits.

   Tools: firecrawl_scrape

   Sources: Firecrawl current MCP tools and troubleshooting, Firecrawl credit billing

4. Keep larger work bounded

   Authenticate before a crawl and give it an explicit scope. Resume a timed-out wait with the existing crawl ID. If the task uses Interact, stop its session when finished.

   Tools: firecrawl_crawl, firecrawl_check_crawl_status, firecrawl_interact_stop

   Sources: Firecrawl current MCP tools and troubleshooting

Selected tools & example tasks

• firecrawl_search

  Find web results from a query; available in keyless mode.

  Example task: Find official documentation URLs before choosing a page.

  Sources: Firecrawl current MCP tools and troubleshooting
• firecrawl_scrape

  Read a known URL; JSON format can extract requested fields. Available in keyless mode.

  Example task: Read a selected public page and return only its needed content.

  Sources: Firecrawl current MCP tools and troubleshooting
• firecrawl_parse

  Parse supported files. Hosted local-file input requires an upload handoff.

  Example task: Convert a document after completing the required upload step.

  Sources: Firecrawl current MCP tools and troubleshooting
• firecrawl_map

  Discover site URLs with authenticated access.

  Example task: Identify a site section before starting a bounded extraction.

  Sources: Firecrawl current MCP tools and troubleshooting
• firecrawl_crawl

  Traverse a site scope with authenticated access; normally waits for a terminal job state.

  Example task: Collect an explicitly limited documentation section.

  Sources: Firecrawl current MCP tools and troubleshooting
• firecrawl_check_crawl_status

  Resume a crawl wait using its job ID.

  Example task: Check an existing job after the original wait times out.

  Sources: Firecrawl current MCP tools and troubleshooting
• firecrawl_interact

  Operate a live page with authenticated access.

  Example task: Navigate or fill a form only within the task you authorized.

  Sources: Firecrawl current MCP tools and troubleshooting
• firecrawl_interact_stop

  Release an Interact session using its scrapeId.

  Example task: Close the browser session when the requested interaction finishes.

  Sources: Firecrawl current MCP tools and troubleshooting

Troubleshoot the connection

Limits & undocumented behavior

• Keyless hosted access exposes Search, Scrape and Parse only; other selected tools require authenticated access and may depend on plan or team policy.
• The current tool guide deprecates the former Extract MCP tool. Use Scrape with JSON for a known page or Agent when sources must be discovered.
• A hosted Parse request cannot directly open a local file. Follow its signed-upload handoff; never assume a local path has already been uploaded.
• The hosted service version and target success rate were not measured. Client-discovered schemas determine the arguments for the connected tool surface.
• A local MCP process is a separate deployment choice and can still call the cloud API. The profiles on this page cover hosted connections.

Sources & review history

This listing is documentation-confirmed. We reviewed the primary sources below; this does not establish successful tool execution.

Record updated 2026-09-28. Review due 2026-10-28.

1. Firecrawl MCP connection modesChecked 2026-09-28 · Hosted keyless, OAuth and API-key modes; plan and team scope
2. Firecrawl keyless and bearer-key configurationChecked 2026-09-28 · Remote endpoint, Cursor examples, keyless tools and bearer header
3. Firecrawl OAuth account connectionChecked 2026-09-28 · OAuth endpoint and client sign-in
4. Firecrawl current MCP tools and troubleshootingChecked 2026-09-28 · Tool names, access scope, Extract deprecation, upload and session behavior
5. Firecrawl keyless and account rate limitsChecked 2026-09-28 · Daily anonymous caps versus team account rate and concurrency limits
6. Firecrawl credit billingChecked 2026-09-28 · Base credits, extraction modifiers and failure exceptions
7. Cursor MCP configuration and authenticationChecked 2026-09-28 · mcpServers JSON, HTTP headers, OAuth and configuration locations
8. Official Firecrawl MCP repositoryChecked 2026-09-28 · Publisher, server implementation and MIT license
9. Firecrawl plans and pricingChecked 2026-09-28 · Account plan offers; consult detailed billing for option charges

Spotted a change? Share the source and the affected setup step in Site Feedback.

Alternatives to Firecrawl MCP

These integrations share documented capabilities with Firecrawl MCP. Compare their setup and product requirements for your task.