Firecrawl MCP server
Firecrawl's official hosted server for scraping, searching and parsing web pages into clean, LLM-ready content. The keyless tier works without authentication at rate limits; crawl, map and agent tools need an account.
Remote (HTTP)No auth
Set up Firecrawl
Configuration examples, not tested connections. Replace placeholders and review the upstream documentation before use.
Claude Code
Run in your terminal.
claude mcp add --transport http firecrawl https://mcp.firecrawl.dev/v2/mcp
Cursor
Merge into ~/.cursor/mcp.json
{
"mcpServers": {
"firecrawl": {
"url": "https://mcp.firecrawl.dev/v2/mcp"
}
}
}
Codex
Run in your terminal.
codex mcp add firecrawl --url https://mcp.firecrawl.dev/v2/mcp
Gemini CLI
Merge into ~/.gemini/settings.json
{
"mcpServers": {
"firecrawl": {
"httpUrl": "https://mcp.firecrawl.dev/v2/mcp"
}
}
}
OpenCode
Merge into ~/.config/opencode/opencode.jsonc
{
"$schema": "https://opencode.ai/config.json",
"mcp": {
"firecrawl": {
"type": "remote",
"url": "https://mcp.firecrawl.dev/v2/mcp",
"enabled": true
}
}
}
VS Code
Merge into .vscode/mcp.json
{
"servers": {
"firecrawl": {
"type": "http",
"url": "https://mcp.firecrawl.dev/v2/mcp"
}
}
}
Using more than one server?Add to a config with other servers
Step-by-step setup guides: Claude Code and Codex CLI, Cursor and VS Code. In Claude Desktop, add the URL as a custom connector. This is a remote server; see remote vs local servers and authentication and security basics.
Overview
Firecrawl's official MCP server gives AI clients web search, page scraping and document parsing. The hosted endpoint is https://mcp.firecrawl.dev/v2/mcp; a local npx -y firecrawl-mcp process is also available. Choose the hosted service for a remote client connection, or local execution when your client needs to launch a process. Installing the MCP package locally does not itself provide a local Firecrawl scraping API.
Keyless hosted access exposes scrape, search and parse within usage limits. OAuth account sign-in and API-key authentication unlock a broader surface according to your team's plan and permissions. Use the client's connection flow for the separate OAuth URL rather than opening it as a website. API-key integrations use the standard URL and a protected Authorization header; keep credentials out of URLs, shared examples and agent chat.
The tool profile matters as much as the URL. Upstream documents 27 tools in the default full profile, including feedback tools, three in the hosted keyless profile, and nine on the separate search-only endpoint. Feedback switches, credentials and deployment settings can change the available list. Check the tools your client actually receives after connecting. Follow Firecrawl MCP setup in VS Code for editor examples; Cloudflare Docs is a separate option for Cloudflare-specific reference searches.
Three ways to connect
| Connection | MCP URL | Authentication |
|---|---|---|
| Keyless | https://mcp.firecrawl.dev/v2/mcp | No key; scrape, search and parse within limits. |
| Account sign-in | https://mcp.firecrawl.dev/v2/mcp-oauth | Client-managed OAuth flow. |
| API key | https://mcp.firecrawl.dev/v2/mcp | Protected Authorization: Bearer <FIRECRAWL_API_KEY> header. |
Run Firecrawl MCP locally
Use Node.js 22 or newer. Provide your real FIRECRAWL_API_KEY through protected client settings. npx -y firecrawl-mcp starts the stdio server; the upstream HTTP example below contains a placeholder key for you to replace locally.
The local client URL is http://localhost:3000/mcp; curl http://localhost:3000/health checks readiness. The hosted /v2/mcp path is different.
env HTTP_STREAMABLE_SERVER=true FIRECRAWL_API_KEY=fc-YOUR_API_KEY npx -y firecrawl-mcp
Which tools do you get?
These are upstream-documented profile counts reviewed on 6 November 2026, not a fresh authenticated tools/list measurement. Check the vendor documentation and your connection for its current list.
| Profile | Upstream count | Qualification |
|---|---|---|
| Hosted keyless | 3 | firecrawl_scrape, firecrawl_search, firecrawl_parse. |
| Default full profile | 27 | Feedback tools included; disabling feedback or using local-keyless startup reduces the count. |
| Search-only | 9 | https://mcp.firecrawl.dev/v2/mcp-search; authenticated search-oriented profile. |
Frequently asked questions
Do I need a Firecrawl MCP API key?
The hosted keyless profile supports scrape, search and parse within limits. Use account sign-in or an API key for broader access, subject to your plan.
Does running npx remove the need for a Firecrawl API?
No. The local MCP process still connects to the Firecrawl cloud API or a configured self-hosted API. Local process execution is a transport choice.
Example tools
firecrawl_scrapefirecrawl_searchfirecrawl_parse
Tool names are upstream examples. The server's live tool list can differ.
Related Developer Tools servers
Sources
This page was written from the upstream documentation below and last reviewed on 6 November 2026. Listing is not a security or compatibility endorsement. Report a correction through the contact page.