MCP Firecrawl
A production-ready Model Context Protocol (MCP) server that integrates with the Firecrawl API to give AI assistants the power to scrape, crawl, and search the web.
README
MCP Firecrawl
A production-ready Model Context Protocol (MCP) server that integrates with the Firecrawl API to give AI assistants the power to scrape, crawl, and search the web.
Works with Claude Desktop, Cursor, VS Code (Cline/Roo Code), Continue, OpenAI Agents SDK, and any other MCP-compatible AI tool.
Features
| Tool | Description | Input |
|---|---|---|
scrape_url |
Scrape any webpage → returns clean Markdown content + metadata | { "url": "https://..." } |
crawl_website |
Crawl multiple pages (up to 100) from a website | { "url": "https://...", "limit": 10 } |
search_web |
Search the web using Firecrawl's search engine | { "query": "latest AI news" } |
Quick Start
# 1. Clone
git clone https://github.com/YOUR_USERNAME/MCP.git
cd MCP
# 2. Install
npm install
# 3. Configure
cp .env.example .env
# Edit .env → add your Firecrawl API key:
# FIRECRAWL_API_KEY=fc-your-key-here
# 4. Build & Run
npm run build
npm start
✅ Server is now running on stdio (default mode).
Usage
Command Line Test
# List available tools
node test-mcp.mjs tools
# Scrape a website
node test-mcp.mjs scrape https://example.com
# Search the web
node test-mcp.mjs search "latest TypeScript news"
With OpenAI Agents SDK (TypeScript)
import { Agent, run, MCPServerStdio } from "@openai/agents";
const mcp = new MCPServerStdio({
name: "Firecrawl",
fullCommand: "node dist/server.js",
env: { FIRECRAWL_API_KEY: "fc-..." },
});
await mcp.connect();
const agent = new Agent({
name: "Web Assistant",
instructions: "Use tools to scrape, crawl, and search the web.",
mcpServers: [mcp],
});
const result = await run(agent, "Scrape https://example.com");
console.log(result.finalOutput);
await mcp.close();
Integration with AI Tools
Claude Desktop
Add to %APPDATA%\Claude\claude_desktop_config.json:
{
"mcpServers": {
"firecrawl": {
"command": "node",
"args": ["C:/path/to/MCP/dist/server.js"],
"env": {
"FIRECRAWL_API_KEY": "fc-..."
}
}
}
}
Cursor
Settings → Features → MCP → Add Server:
Name: firecrawl
Type: command
Command: node "C:/path/to/MCP/dist/server.js"
Environment: FIRECRAWL_API_KEY=fc-...
VS Code + Cline / Roo Code
File: .vscode/cline_mcp_settings.json
{
"mcpServers": {
"firecrawl": {
"command": "node",
"args": ["C:/path/to/MCP/dist/server.js"],
"env": { "FIRECRAWL_API_KEY": "fc-..." },
"disabled": false,
"autoApprove": []
}
}
}
Continue (VS Code / JetBrains)
File: ~/.continue/config.json
{
"experimental": {
"mcpServers": {
"firecrawl": {
"command": "node",
"args": ["C:/path/to/MCP/dist/server.js"],
"env": { "FIRECRAWL_API_KEY": "fc-..." }
}
}
}
}
opencode
File: ./opencode.json
{
"mcp": {
"firecrawl": {
"type": "local",
"command": ["node", "MCP/dist/server.js"],
"enabled": true,
"env": { "FIRECRAWL_API_KEY": "fc-..." }
}
}
}
HTTP Mode (Streamable HTTP)
Can't use stdio? Run the server as an HTTP endpoint instead:
npm run start:http
# Listening on http://127.0.0.1:3001/mcp
Then connect any MCP client to that URL.
Project Structure
MCP/
├── src/
│ ├── server.ts # Entry point (stdio + HTTP modes)
│ ├── tools/
│ │ ├── scrape.ts # scrape_url tool
│ │ ├── crawl.ts # crawl_website tool
│ │ └── search.ts # search_web tool
│ ├── services/
│ │ └── firecrawl.ts # FirecrawlApp API wrapper
│ └── types/
│ └── index.ts # TypeScript interfaces
├── test-mcp.mjs # CLI test utility
├── .env.example # Environment template
├── package.json
├── tsconfig.json
└── README.md
Environment Variables
| Variable | Required | Default | Description |
|---|---|---|---|
FIRECRAWL_API_KEY |
✅ Yes | — | Your Firecrawl API key |
HTTP_HOST |
❌ No | 127.0.0.1 |
HTTP bind address (HTTP mode only) |
HTTP_PORT |
❌ No | 3001 |
HTTP port (HTTP mode only) |
NPM Scripts
| Script | Description |
|---|---|
npm run build |
Compile TypeScript → dist/ |
npm start |
Run in stdio mode |
npm run start:http |
Run in HTTP mode |
npm run dev |
Run with hot-reload (stdio) |
npm run dev:http |
Run with hot-reload (HTTP) |
npm run clean |
Delete dist/ |
Tech Stack
- Runtime: Node.js 18+
- Language: TypeScript
- MCP SDK:
@modelcontextprotocol/sdk - Scraping Engine:
firecrawlSDK - Validation:
zod - Config:
dotenv
Requirements
- Node.js 18 or higher
- A Firecrawl API key (free tier available)
License
MIT
推荐服务器
Baidu Map
百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。
Playwright MCP Server
一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。
Magic Component Platform (MCP)
一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。
Audiense Insights MCP Server
通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。
VeyraX
一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。
graphlit-mcp-server
模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。
Kagi MCP Server
一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。
e2b-mcp-server
使用 MCP 通过 e2b 运行代码。
Neon MCP Server
用于与 Neon 管理 API 和数据库交互的 MCP 服务器
Exa MCP Server
模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。