CrewAI

CrewAI agents can use Search1API through the hosted MCP server at https://mcp.search1api.com/mcp. No extra package is needed: CrewAI's mcps field connects to the server, discovers its tools, and hands them to the agent. You get web search, news, agentic search, and page reading from one connection.

InstallCopy Anchor Link

pip install crewai mcp

Create a key in the dashboard, then export it:

export SEARCH1API_API_KEY="your-api-key"

Add Search1API to an agentCopy Anchor Link

Pass an MCPServerHTTP configuration with the key in the Authorization header. The tool filter keeps the agent to the tools a research crew needs.

import os

from crewai import Agent, Crew, Task
from crewai.mcp import MCPServerHTTP
from crewai.mcp.filters import create_static_tool_filter

search1api = MCPServerHTTP(
    url="https://mcp.search1api.com/mcp",
    headers={"Authorization": f"Bearer {os.environ['SEARCH1API_API_KEY']}"},
    tool_filter=create_static_tool_filter(
        allowed_tool_names=["search", "news", "ask", "crawl"]
    ),
    cache_tools_list=True,
)

researcher = Agent(
    role="Research Analyst",
    goal="Answer questions with current, cited sources from the web",
    backstory="You search first, read the most relevant pages, and cite every claim.",
    mcps=[search1api],
)

task = Task(
    description="What changed in CrewAI this month? Read the release notes before answering.",
    expected_output="A short summary with a source URL for each change",
    agent=researcher,
)

result = Crew(agents=[researcher], tasks=[task]).kickoff()
print(result)

The agent picks a model the usual CrewAI way, from llm= or your environment.

ToolsCopy Anchor Link

The server exposes the same tools to CrewAI as to any MCP client. The ones most useful for agents:

maps and trending are also available. See the MCP server page for every tool and its inputs, and Credits and limits for current rates.

Two tools are meant for chat apps rather than agents: render_results draws source cards in a client UI, and send_feedback reports problems to the Search1API team. Leave them out of allowed_tool_names for crews.

Quick setup with a URLCopy Anchor Link

For a quick test, CrewAI also accepts the server as a string. Add #search to load only one tool:

researcher = Agent(
    role="Research Analyst",
    goal="Answer questions with current sources",
    backstory="You search before you answer.",
    mcps=[f"https://mcp.search1api.com/mcp?apiKey={os.environ['SEARCH1API_API_KEY']}#search"],
)

This puts the key in a URL, where it can end up in logs. Use the header configuration above for anything you deploy.

CrewBase projectsCopy Anchor Link

In a project created with crewai create crew, set mcps on the agent inside its @agent method:

@agent
def researcher(self) -> Agent:
    return Agent(config=self.agents_config["researcher"], mcps=[search1api])

RelatedCopy Anchor Link

• MCP server: authentication, OAuth, and every tool
• LangChain: native tools for LangChain agents
• CrewAI MCP documentation