from langchain_mcp_adapters.client import MultiServerMCPClient
from langgraph.prebuilt import create_react_agent
from langchain_cerebras import ChatCerebras
import asyncio
from dotenv import load_dotenv
load_dotenv()
async def main():
# Setup MCP client config
client = MultiServerMCPClient({
"agenthotspot": {
"transport": "stdio",
"command": "python3",
"args": ["-m", "agenthotspot_mcp"],
"env": {}
}
})
# Initialize LLM and agent
tools = await client.get_tools()
# Replace with llm client of choice
# Make sure to set CEREBRAS_API_KEY=... in .env environment file.
llm = ChatCerebras(model="gpt-oss-120b", temperature=0.6)
agent = create_react_agent(llm, tools)
# Invoke agent and output response with insights
response = await agent.ainvoke({
"messages": [{"role": "user", "content": "Find memory related MCPs."}],
})
print(response["messages"][-1].content)
if __name__ == "__main__":
asyncio.run(main())
