langgraphlisted

# LangGraph Skill ## When to activate - Building a stateful agent that needs to pause, resume, or branch based on conditions - Implementing human-in-the-loop workflows (agent pauses for approval before continuing) - Orchestrating multi-step agents with complex branching logic - Persisting agent state across failures or sessions (checkpointing) - Building multi-agent systems where subgraphs collaborate ## When NOT to use - Simple single-turn AI calls — use the Claude API or Vercel AI SDK skill directly - Linear pipelines with no branching — overkill, use sequential function calls - When you need a hosted agent loop — use the Mastra skill instead - Frontend streaming chat UIs — use the Vercel AI SDK skill ## Instructions ### Installation ```bash npm install @langchain/langgraph @langchain/anthropic @langchain/core # Python pip install langgraph langchain-anthropic ``` ### Core concepts LangGraph models agents as **graphs**: - **Nodes** — functions that run and update state - **Edges** — connections between nodes (always run → next node, or conditional) - **State** — a typed object that flows through the graph and accumulates updates - **Checkpointer** — persists state between runs (enables resume, human-in-the-loop) ### Basic StateGraph (TypeScript) ```typescript import { StateGraph, Annotation, END, START } from '@langchain/langgraph' import { ChatAnthropic } from '@langchain/anthropic' import { tool } from '@langchain/core/tools' import { z } from 'zod' // 1. Define