firecrawl-toolkit 0.0.8

A high-performance, asynchronous MCP server for Firecrawl Search, featuring connection pooling, request retries, and intelligent input parsing.

Firecrawl MCP Toolkit

A high-performance, asynchronous MCP server that provides comprehensive Google search and web content scraping capabilities through the Firecrawl API (excluding some rarely used interfaces).

This project is built on httpx, utilizing asynchronous clients and connection pool management to offer LLMs a stable and efficient external information retrieval tool.

Key Features

• Asynchronous Architecture: Fully based on asyncio and httpx, ensuring high throughput and non-blocking I/O operations.
• HTTP Connection Pool: Manages and reuses TCP connections through a global httpx.AsyncClient instance, significantly improving performance under high concurrency.
• Concurrency Control: Built-in global and per-API endpoint concurrency semaphores effectively manage API request rates to prevent exceeding rate limits.
• Automatic Retry Mechanism: Integrated request retry functionality with exponential backoff strategy automatically handles temporary network fluctuations or server errors, enhancing service stability.
• Intelligent Country Code Parsing: Includes a comprehensive country name dictionary supporting inputs in Chinese, English, ISO Alpha-2/3, and other formats, with automatic normalization.
• Flexible Environment Variable Configuration: Supports fine-tuned service configuration via environment variables.

Available Tools

This service provides the following tools:

Tool Name	Description
firecrawl-search	Performs general Google web / news / images searches.
firecrawl-scrape	Scrapes and returns the content of a specified URL.

Installation Guide

It is recommended to install using pip or uv.

# Using pip
pip install firecrawl-toolkit

# Or using uv
uv pip install firecrawl-toolkit

Quick Start

Set Environment Variables

Create a .env file in the project root directory and enter your Firecrawl API key:

FIRECRAWL_API_KEY="your-firecrawl-api-key-here"
FIRECRAWL_HTTP2=<0/1>
FIRECRAWL_MAX_WORKERS=<int, default 10>
FIRECRAWL_MAX_CONNECTIONS=<int, default 200>
FIRECRAWL_MAX_CONCURRENT_REQUESTS=<int, default 200>
FIRECRAWL_KEEPALIVE=<int, default 20>
FIRECRAWL_RETRY_COUNT=<int, default 3>
FIRECRAWL_MCP_ENABLE_STDIO=<0/1>
FIRECRAWL_MCP_ENABLE_HTTP=<0/1>
FIRECRAWL_MCP_ENABLE_SSE=<0/1>
FIRECRAWL_MCP_HTTP_HOST=<str, 127.0.0.1>
FIRECRAWL_MCP_HTTP_PORT=<str, 7001>
FIRECRAWL_MCP_SSE_HOST=<str, 127.0.0.1>
FIRECRAWL_MCP_SSE_PORT=<str, 7001>

Configure MCP Client

Add the following server configuration in the MCP client configuration file:

{
  "mcpServers": {
    "firecrawl": {
      "command": "python3",
      "args": ["-m", "firecrawl-toolkit"],
      "env": {
        "FIRECRAWL_API_KEY": "<Your Firecrawl API key>"
      }
    }
  }
}

{
  "mcpServers": {
    "firecrawl": {
      "command": "uvx",
      "args": ["firecrawl-toolkit"],
      "env": {
        "FIRECRAWL_API_KEY": "<Your Firecrawl API key>"
      }
    }
  }
}

Environment Variables

• FIRECRAWL_MAX_CONNECTIONS: Maximum number of HTTP client connections (default: 200).
• FIRECRAWL_KEEPALIVE: Maximum number of keep-alive HTTP client connections (default: 20).
• FIRECRAWL_HTTP2: Enable HTTP/2 (default: "0", set to "1" to enable).
• FIRECRAWL_MAX_CONCURRENT_REQUESTS: Global maximum concurrent requests (default: 200).
• FIRECRAWL_RETRY_COUNT: Maximum retry attempts for failed requests (default: 3).
• FIRECRAWL_RETRY_BASE_DELAY: Base delay time for retries in seconds (default: 0.5).
• FIRECRAWL_ENDPOINT_CONCURRENCY: Set concurrency per endpoint (JSON format), e.g., {"search":10,"scrape":2}.
• FIRECRAWL_ENDPOINT_RETRYABLE: Set retry allowance per endpoint (JSON format), e.g., {"scrape": false}.

Tool Parameters and Usage Examples

firecrawl-search: Perform web / news / images search

Parameters:

• query (str, required): Keywords to search.
• country (str, optional): Specify the country/region for search results. Supports Chinese names (e.g., "China"), English names (e.g., "United States"), or ISO codes (e.g., "US"). Default is "US".
• search_num (int, optional): Number of results to return, range 1-100. Default is 20.
• search_time (str, optional): Filter results by time range. Available values: "hour", "day", "week", "month", "year".

Example:

result_json = firecrawl_search(
    query="AI advancements 2024",
    country="United States",
    search_num=5,
    search_time="month"
)

firecrawl-scrape: Scrape webpage content

Parameters:

• url (str, required): URL of the target webpage.

Example:

result_json = firecrawl_scrape(
    url="https://www.example.com"
)

License Agreement

This project is licensed under the MIT License.