# Checkpointing

LangGraph saves the state after every step (in memory or Postgres). A conversation can resume tomorrow, and a crash doesn't lose progress.

AI developers · Part 06: LangChain / LangGraph · Card 12 of 14 · Last updated 2026-10-13

## Analogy

Autosave in a game. Die at the boss and you restart at the last save point, not the title screen.

## Code example

`thread.ts`

```typescript
const config = { configurable: { thread_id: "user-42" } };

await graph.invoke({ messages: [{ role: "user", content: "I'm Faisal" }] }, config);
await graph.invoke({ messages: [{ role: "user", content: "Who am I?" }] }, config);
// same thread_id -> the graph remembers
```

## In an agent

Checkpoints give an agent working memory for multi-step tasks.

## Practice

### Flashcards

- **What does a checkpointer do?** Saves agent state after each step so a run can resume later (LangGraph).
- **Interview: three kinds of agent memory?** Short-term message history, long-term facts in a database, and working state in a checkpointer.
- **Interview: what does LangGraph add to LangChain?** Explicit graphs with state, checkpoints, interrupts and routing, so complex agents are easier to control.

### Interview questions

**How do you handle memory?**

- Short-term memory is the message history, trimmed or summarized near the context limit.
- Long-term facts go to a database and are retrieved when relevant.
- For multi-step tasks, LangGraph checkpoints in Postgres hold the working state.

## Questions people ask

### What is checkpointing in LangGraph?

Checkpointing saves the graph's state after every step, in memory for development or in a database such as Postgres for production. Each conversation uses a thread_id. With the same thread_id, a conversation can resume tomorrow, and a crash in the middle of a task does not lose the progress already made.

### How do LangGraph checkpoints give an agent memory?

Pass the same thread_id in the config on every call. LangGraph loads that thread's saved state, including the earlier messages, before running. So if a user says their name in one call, the agent still knows it in the next one, even days later, because the state was stored.

## Related concepts

- [Short-term, long-term and working memory](https://studyvisually.com/learn/tech/ai-developers/production/memory-types)
- [LangGraph.js](https://studyvisually.com/learn/tech/ai-developers/langchain-langgraph/langgraph)
- [Graph state](https://studyvisually.com/learn/tech/ai-developers/langchain-langgraph/graph-state)
- [Conditional edges](https://studyvisually.com/learn/tech/ai-developers/langchain-langgraph/conditional-edges)

Web page: https://studyvisually.com/learn/tech/ai-developers/langchain-langgraph/checkpointing
Study this card in the app: https://studyvisually.com/app/learn?category=ai-developers&card=checkpointing