langchain-agents

Build LLM agents with LangChain by defining tool-calling abstractions that let models invoke external functions. LangGraph manages stateful execution, enabling complex multi-turn interactions where agents research, summarize, and act sequentially.

Switch between OpenAI and Anthropic models without altering core agent logic using provider-agnostic abstractions. LangChain standardizes the interface so tool integration and state management remain consistent across different LLM providers.

Debug LLM agent execution and track costs using LangSmith observability. It provides detailed tracing and analysis of agent behavior, allowing you to inspect tool calls, monitor state transitions, and identify errors in complex execution graphs.

Add human-in-the-loop approval to stateful LangGraph agents by defining complex execution graphs with persistent state management. LangGraph facilitates pausing execution for human review before proceeding with sensitive tool calls or actions.

LangGraph supports multi-turn interactions for agents requiring persistent state. It enables stateful agent execution by maintaining context across turns, allowing agents to handle complex workflows like researching topics and drafting emails sequentially.

Use LangGraph for stateful agents needing complex execution graphs and human-in-the-loop control. Standard LangChain chains lack persistent state management across multi-turn interactions, while LangGraph maintains context and enables intricate agent workflows.