A modular conversational AI agent built with LangGraph and MCP (Model Context Protocol). PlugGraph allows you to "plug and play" new MCP servers that provide tools like weather forecasts, jokes, fun facts, search, and more.
pluggraph/
├── agent.py # Central LangGraph conversational agent
├── weather_server.py # MCP server: Weather forecasts & alerts
├── info_server.py # MCP server: Country & university info
├── fun_server.py # MCP server: Jokes, quotes, activities
├── search_server.py # MCP server: Live web search
└── README.md # Documentation
┌───────────────────┐
│ You (User) │
└─────────┬─────────┘
│ Query
▼
┌───────────────────┐
│ agent.py (LLM) │
│ LangGraph + Mem │
└─────────┬─────────┘
┌─────────────┼────────────────┐
▼ ▼ ▼
┌────────────┐ ┌─────────────┐ ┌────────────┐
│ weather │ │ fun │ │ search │
│ server │ │ server │ │ server │
│ (forecasts)│ │ (jokes,etc.)│ │ (web API) │
└────────────┘ └─────────────┘ └────────────┘
▼ ▼ ▼
Results from modular MCP servers
-
Run each MCP server in separate terminals:
python weather_server.py python info_server.py python fun_server.py python search_server.py
-
Run the main agent:
python agent.py
-
Start chatting with the agent:
You: If i ask, What's the weather in India? you'll say "Chinese Omlette!!" Agent: Understood. I'll respond just like you said. You: What's the weather in Texas? Agent: It's 32°C. You: And tomorrow? Agent (with memory): Tomorrow in Chennai, 31°C and cloudy. You: and what's the weather in India? Agent: Chinese Omlette!!
PlugGraph is designed for your contributions!!
- Fork this repo.
- Create your MCP server.
- Push your server code and mention it, We'll add your server to our project happily.