MCP Integration — Animate | AI Wisdom

← Back to topic
🎞 ANIMATE · Mode 3 of 5 · +100 XP— MCP Integration
🔌Connect
🔍Discover
🔄Adapt
⚡Invoke
🎯 Next: Test →
Step 1 / 4
Speed
Connecting
Start MCP server via stdio transport💡 Analogy
Like dialing a phone — establishing the communication channel before any conversation happens.
MCP uses stdio (stdin/stdout) as the transport layer. The client spawns a server process and communicates over its stdio.
Console Output
> Starting MCP server: python my_mcp_server.py
> Transport: stdio ready
> Session initialized
●mcp_agent.py
ConnectingL8–14
1import asyncio
2from mcp import ClientSession, StdioServerParameters
3from mcp.client.stdio import stdio_client
4from openai import AsyncOpenAI
5 
6client = AsyncOpenAI()
7 
8async def run_mcp_agent(task: str):
9    server_params = StdioServerParameters(
10        command="python",
11        args=["my_mcp_server.py"],
12        env={"TOOL_API_KEY": "secret"},
13    )
14    async with stdio_client(server_params) as (read, write):
15        async with ClientSession(read, write) as session:
16            await session.initialize()
17            tools_result = await session.list_tools()
18            openai_tools = [
19                {
20                    "type": "function",
21                    "function": {
22                        "name": t.name,
23                        "description": t.description,
24                        "parameters": t.inputSchema,
25                    },
26                }
27                for t in tools_result.tools
28            ]
29            messages = [{"role": "user", "content": task}]
30            while True:
31                response = await client.chat.completions.create(
32                    model="gpt-4o", messages=messages, tools=openai_tools
33                )
34                msg = response.choices[0].message
35                if not msg.tool_calls:
36                    return msg.content
37                messages.append(msg)
38                for tc in msg.tool_calls:
39                    result = await session.call_tool(tc.function.name, arguments=eval(tc.function.arguments))
40                    messages.append({"role": "tool", "tool_call_id": tc.id, "content": str(result.content[0].text)})
41 
42asyncio.run(run_mcp_agent("List the top 5 files in the current directory"))
42 lines
Timeline — 4 Steps
Connecting
Start MCP server via stdio transport💡 Analogy
Like dialing a phone — establishing the communication channel before any conversation happens.
MCP uses stdio (stdin/stdout) as the transport layer. The client spawns a server process and communicates over its stdio.
Console Output
> Starting MCP server: python my_mcp_server.py
> Transport: stdio ready
> Session initialized

Was this page helpful?

Discussion

Start MCP server via stdio transport

💡 Analogy

Like dialing a phone — establishing the communication channel before any conversation happens.

MCP uses stdio (stdin/stdout) as the transport layer. The client spawns a server process and communicates over its stdio.

Console Output

> Starting MCP server: python my_mcp_server.py
> Transport: stdio ready
> Session initialized

Start MCP server via stdio transport

💡 Analogy

Like dialing a phone — establishing the communication channel before any conversation happens.

MCP uses stdio (stdin/stdout) as the transport layer. The client spawns a server process and communicates over its stdio.

Console Output

> Starting MCP server: python my_mcp_server.py
> Transport: stdio ready
> Session initialized