[AI] MCP Utilization: LangChain & LangGraph

LangChain

• A framework for developing applications based on large language models (LLMs).
• Features
  • Connects various LLMs, Tools, and Memory in chain format.
  • Supports integration with external data sources.
  • Linear chain structure that executes in predefined workflow order.
• Differences from existing LLMs
  • While existing LLMs focus on text generation and question & answer, LangChain enables building complex and practical applications by integrating with external data.

LangGraph

• A framework that enables implementing flexible and dynamic workflows using graph structures.
• Core Components
  • Node
    • Individual work units executed in the workflow.
    • Can execute functions such as LLM calls, tool usage, and data processing.
  • Edge
    • Connections between nodes that determine the workflow.
  • State
    • A data storage shared by all nodes.
    • Multiple agents share information through state and maintain workflow context.
  • Agent
    • LLM can perform the 'agent' role of determining which node to execute.
• Features
  • Can internally track state and freely call appropriate MCPs.