MCP Integration¶
AgentiCraft seamlessly integrates with the Model Context Protocol (MCP) for enhanced tool capabilities.
Overview¶
MCP enables agents to use tools from any MCP-compatible server, expanding capabilities without custom code.
Basic Usage¶
from agenticraft import Agent
from agenticraft.mcp import MCPClient
# Connect to MCP server
mcp = MCPClient("http://localhost:8080")
# Create agent with MCP tools
agent = Agent(
name="MCPAgent",
model="gpt-4",
tools=mcp.get_tools()
)
# Use MCP tools naturally
response = agent.run("Search for the latest AI news")
Available MCP Servers¶
File System Tools¶
mcp = MCPClient("mcp://filesystem")
agent = Agent(name="FileBot", tools=mcp.get_tools())
agent.run("List all Python files in the current directory")
agent.run("Read the README.md file")
Database Tools¶
mcp = MCPClient("mcp://postgres",
connection_string="postgresql://localhost/mydb"
)
agent = Agent(name="DataBot", tools=mcp.get_tools())
agent.run("Show me all users created this month")
Web Tools¶
mcp = MCPClient("mcp://web")
agent = Agent(name="WebBot", tools=mcp.get_tools())
agent.run("Search for AgentiCraft tutorials")
agent.run("Get the content of https://example.com")
Custom MCP Servers¶
from agenticraft.mcp import MCPServer, mcp_tool
class CustomMCPServer(MCPServer):
@mcp_tool
def get_weather(self, location: str) -> str:
"""Get weather for a location."""
# Implementation
return f"Sunny in {location}"
# Start server
server = CustomMCPServer()
server.start(port=8080)
# Use in agent
mcp = MCPClient("http://localhost:8080")
agent = Agent(name="WeatherBot", tools=mcp.get_tools())
Best Practices¶
-
Tool Discovery
-
Error Handling
-
Performance
- Use connection pooling for high-throughput
- Cache tool definitions
- Implement timeouts for reliability