Agent Development Kit (ADK)

Integrate Firecrawl with Google's Agent Development Kit (ADK) to build powerful AI agents with web scraping capabilities through the Model Context Protocol (MCP).

Overview#

Firecrawl provides an MCP server that seamlessly integrates with Google's ADK, enabling your agents to efficiently scrape, crawl, and extract structured data from any website. The integration supports both cloud-based and self-hosted Firecrawl instances with streamable HTTP for optimal performance.

Features#

• Efficient web scraping, crawling, and content discovery from any website
• Advanced search capabilities and intelligent content extraction
• Deep research
• Flexible deployment (cloud-based or self-hosted)
• Optimized for modern web environments with streamable HTTP support

Prerequisites#

• Obtain an API key for Firecrawl from firecrawl.dev
• Install Google ADK

Setup#

import osfrom google.adk.agents.llm_agent import Agentfrom google.adk.tools.mcp_tool.mcp_session_manager import StreamableHTTPServerParamsfrom google.adk.tools.mcp_tool.mcp_toolset import MCPToolsetFIRECRAWL_API_KEY = os.environ["FIRECRAWL_API_KEY"]root_agent = Agent(    model="gemini-2.5-pro",    name="firecrawl_agent",    description='A helpful assistant for scraping websites with Firecrawl',    instruction='Help the user search for website content',    tools=[        MCPToolset(            connection_params=StreamableHTTPServerParams(                url="https://mcp.firecrawl.dev/v2/mcp",                headers={"Authorization": f"Bearer {FIRECRAWL_API_KEY}"},            ),        )    ],)

Set FIRECRAWL_API_KEY in the environment that starts ADK. Do not put its value in agent prompts, source control, or the MCP URL. The local MCP package requires Node.js 22 or newer.

Available Tools#

Configuration#

Required Configuration#

FIRECRAWL_API_KEY: Your Firecrawl API key

• Required when using cloud API (default)
• Optional when using self-hosted instance with FIRECRAWL_API_URL

Optional Configuration#

Firecrawl API URL (for self-hosted instances):

• FIRECRAWL_API_URL: Custom API endpoint
• Example: https://firecrawl.your-domain.com
• If not provided, the cloud API will be used

Example: Web Research Agent#

from google.adk.agents.llm_agent import Agentfrom google.adk.tools.mcp_tool.mcp_session_manager import StreamableHTTPServerParamsfrom google.adk.tools.mcp_tool.mcp_toolset import MCPToolsetFIRECRAWL_API_KEY = "YOUR-API-KEY"# Create a research agentresearch_agent = Agent(    model="gemini-2.5-pro",    name="research_agent",    description='An AI agent that researches topics by scraping and analyzing web content',    instruction='''You are a research assistant. When given a topic or question:    1. Use the search tool to find relevant websites    2. Scrape the most relevant pages for detailed information    3. Extract structured data when needed    4. Provide comprehensive, well-sourced answers''',    tools=[        MCPToolset(            connection_params=StreamableHTTPServerParams(                url="https://mcp.firecrawl.dev/v2/mcp",                headers={"Authorization": f"Bearer {FIRECRAWL_API_KEY}"},            ),        )    ],)# Use the agentresponse = research_agent.run("What are the latest features in Python 3.13?")print(response)

Best Practices#

1. Use the right tool for the job:

   • firecrawl_search when you need to find relevant pages first
   • firecrawl_scrape for single pages
   • firecrawl_crawl for discovering and scraping entire sites
   • repeated firecrawl_scrape calls when you already have a short list of known URLs

2. Monitor your usage: Use your Firecrawl dashboard and API responses to track credit usage.

3. Handle errors gracefully: Surface MCP/API errors to the user and retry only when your agent workflow can do so safely.

4. Optimize performance: Use firecrawl_map before scraping when the agent needs to discover relevant URLs.

Related Resources#