Connect MCP to the OpenAI Agents SDK

The OpenAI Agents SDK supports multiple MCP integration styles. Use hosted MCP when the Responses API should call a publicly reachable server on your behalf; use Streamable HTTP when your application manages the connection; use stdio for a local subprocess.

Streamable HTTP example

from agents import Agent, Runner
from agents.mcp import MCPServerStreamableHttp

async with MCPServerStreamableHttp(
    name="Internal MCP",
    params={"url":"https://mcp.example.com/mcp"},
    cache_tools_list=True,
) as server:
    agent = Agent(
        name="Assistant",
        instructions="Use approved tools when needed.",
        mcp_servers=[server],
    )
    result = await Runner.run(agent, "Find order A-100")

Choose the integration boundary deliberately

Hosted MCP reduces local round-trip orchestration. Locally managed MCP gives your code direct control over headers, filtering, metadata, and connection lifecycle.

Reference

Official MCP integration guide