A hands-on learning repository covering LangGraph fundamentals — from basic chatbots and human-in-the-loop workflows to multimodal RAG, multi-agent systems, and MCP server integration with LangChain.
langgraph-mcp-agents
A hands-on learning repository covering LangGraph fundamentals — from basic chatbots and human-in-the-loop workflows to multimodal RAG, multi-agent systems, and MCP server integration with LangChain.
📁 Project Structure 2-LANGGRPAH-MCP/ ├── 1-BasicChatbot/ │ └── 1-basicchatbot.ipynb # Simple LangGraph chatbot with message state ├── 2-HumanAssistance/ │ └── humaninloop.ipynb # Human-in-the-loop with interrupt and approval flows ├── 3-Debugging/ │ ├── agent.py # Agent definition for debugging demos │ ├── debugging.ipynb # LangGraph Studio & LangSmith tracing walkthrough │ ├── LangGraph_Debug_Monitoring_Notes.docx │ └── langgraph.json # LangGraph Studio config ├── 4-Multimodal/ │ ├── multimodal_rag_notes.docx │ ├── multimodal_sample.pdf │ └── multimodalopenai.ipynb # Multimodal RAG with vision + OpenAI ├── Agents/ │ ├── multi_agent_langgraph_notes.docx │ └── multiaiagent.ipynb # Multi-agent orchestration with LangGraph ├── mcpdemolangchain/ │ ├── client.py # MCP client connecting to tool servers │ ├── mathserver.py # MCP math tool server │ └── weather.py # MCP weather tool server ├── main.py # Entry point ├── pyproject.toml ├── requirements.txt └── .env # API keys (not committed)
🧩 Modules Overview 1 — Basic Chatbot Build a minimal stateful chatbot using StateGraph with MessagesState. Covers nodes, edges, and how LangGraph manages conversation history. 2 — Human Assistance (Human-in-the-Loop) Implement interrupt() to pause graph execution for human review or approval. Demonstrates how to resume the graph after human input using Command. 3 — Debugging & Monitoring Use LangGraph Studio for visual graph inspection and LangSmith for tracing. Covers how to structure a langgraph.json config, set checkpointers, and monitor state transitions. 4 — Multimodal RAG Build a Retrieval-Augmented Generation pipeline that handles both text and images using OpenAI's vision models. Includes a sample PDF and notes. 5 — Multi-Agent Systems Orchestrate multiple AI agents using LangGraph's supervisor and handoff patterns. Each agent has a specialized role; the graph routes tasks between them. 6 — MCP Demo (LangChain + MCP) Connect LangChain agents to external tools via the Model Context Protocol (MCP). Includes a local math server and weather server, and a client that wires them together.
🛠️ Tech Stack ToolPurposeLangGraphStateful agent graph frameworkLangChainLLM chains and tool integrationsOpenAILLM + vision model providerLangSmithTracing and observabilityMCPModel Context Protocol for tool servers
📚 Learning Path Follow the modules in order for the best experience: 1-BasicChatbot → 2-HumanAssistance → 3-Debugging → 4-Multimodal → Agents → MCP