MCP AI Infra Real Time Agent
@junfanz1
This project demonstrates a decoupled real-time agent architecture that connects LangGraph agents to remote tools served by custom MCP (Modular Command Protocol) servers. The architecture enables a flexible and scalable multi-agent system where each tool can be hosted independent
Overview
What is MCP AI Infra Real Time Agent?
A decoupled real-time agent architecture that connects LangGraph agents to remote tools hosted by custom MCP (Model Context Protocol) servers. It enables LLM orchestration to be separated from tool execution, supporting both SSE and STDIO transports for modular, cloud-deployable multi-agent systems.
How to use MCP AI Infra Real Time Agent?
Run the provided MCP tool servers (e.g., python math_server.py, python weather_server.py) then execute a LangGraph agent client such as multiserver_client.py or stdio_client.py. Configure API keys via a .env file and use the MultiServerMCPClient to bind to multiple tool servers concurrently. Debug tool execution with Anthropic’s MCP Inspector.
Key features of MCP AI Infra Real Time Agent
- Decoupled architecture for agent orchestration and tool execution.
- Supports both SSE and STDIO transport protocols.
- Asynchronous I/O for concurrent and real-time communication.
- MultiServerMCPClient enables connections to multiple tool servers.
- Dynamic tool discovery and MCP handshake protocol.
- LangGraph ReAct agent integration for reasoning and acting.
Use cases of MCP AI Infra Real Time Agent
- Building multi-agent systems where agents invoke remote tools from separate servers.
- Real-time weather and math tool execution for AI assistants.
- Integrating LangGraph‑based agents with MCP‑compatible tool servers.
- Developing scalable, cloud‑deployable AI infrastructure with modular tool execution.
FAQ from MCP AI Infra Real Time Agent
What transports are supported for tool servers?
The system supports both SSE (Server‑Sent Events) and STDIO
