MCPBus: Turn any OpenAPI Spec into an MCP Server in 1 Second.

Claude Desktop

Open Claude Desktop config file claude_desktop_config.json (can be accessed via Claude Desktop UI in Settings->Developer->Edit Config).

In the config "mcpServers" section - add JSON entry describing your MCP server which was run with MCPBus:

{
  "mcpServers": {
    "My MCP server": {
      "command": "/usr/local/bin/npx",
      "args": [
        "mcp-remote",
        "http://localhost:8080/mcp",
        "--transport",
        "http-first",
        "--debug"
      ]
    }
  }
}

Restart Claude Desktop.

OpenAI

Example of how to connect your AI-agent (with OpenAI Agents SDK) to your MCP-server which was run with MCPBus:

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

async def main():
    server_params = {"url": "http://localhost:8080/mcp"}

    async with MCPServerStreamableHttp(
            name="MyLocalServer", params=server_params) as mcp_server:
        
        agent = Agent(
            name="LocalMcpAgent",
            instructions="You are a helpful weather expert.",
            mcp_servers=[mcp_server]
        )
        
        result = await Runner.run(
            agent, input="Can you check weather using my local tool?")
        print(result.text)

if __name__ == "__main__":
    asyncio.run(main())

LangChain

First, install the adapters library along with LangGraph and your preferred LLM provider (e.g., OpenAI):

pip install langchain-mcp-adapters langgraph langchain-openai

Use this code example to connect to locally running MCPBus server:

import asyncio
from langchain_openai import ChatOpenAI
from langchain_mcp_adapters.client import MultiServerMCPClient
from langgraph.prebuilt import create_react_agent

async def main():
    client = MultiServerMCPClient(
        {
            "my_local_service": {
                "transport": "http",
                "url": "http://localhost:8080/mcp"
            }
        }
    )
    tools = await client.get_tools()    
    model = ChatOpenAI(model="gpt-4o")
    agent = create_react_agent(model, tools)
    response = await agent.ainvoke(
        {
            "messages": [
                {
                    "role": "user",
                    "content": "Fetch data using my local server tools"
                }
            ]
        }
    )
    print(response["messages"][-1].content)

if __name__ == "__main__":
    asyncio.run(main())