What is LangGraph?
Updated on
LangGraph is a stable 1.x low-level orchestration runtime for long-running, stateful workflows that combine deterministic code, model calls, tools, and human input.
Quick Facts
| Full Name | LangGraph Agent Orchestration Runtime |
|---|---|
| Created | Launched by the LangChain team to solve the pain point of traditional Chains lacking loop capabilities |
| Specification | Official Specification |
How It Works
State, checkpoints, and stores have different scopes
Nodes return partial updates, and each state channel's reducer determines how updates combine. A checkpointer records thread-scoped snapshots for continuity, interrupts, replay, and recovery; a store holds application data across threads. Compiling a graph is not enough: restart recovery requires a stable thread ID and a persistent checkpointer. In-memory savers disappear with the process.
Durability depends on deterministic replay
A resumed run starts from a node, subgraph boundary, or functional entrypoint rather than the exact source line. Put non-deterministic work and side effects in explicit tasks or nodes, use stable operation and idempotency keys, and reconcile unknown outcomes. Time travel creates a replay or fork; it cannot reverse an email, payment, deployment, or database write already committed.
Graceful drain is cooperative, not cancellation
LangGraph 1.2 adds RunControl-based graceful shutdown between supersteps. A running node or retry loop finishes before the drain takes effect, a resumable checkpoint is saved, and the run can continue with the same thread identity. Drain does not cancel in-flight async tasks, so deployments still need a bounded shutdown deadline and downstream cancellation or idempotency policy.
Key Characteristics
- Provides a low-level orchestration Runtime rather than prescribing one Agent, Prompt, Tool, or Multi-Agent architecture
- Supports declarative Graph and Functional APIs on the same persistence, streaming, and interrupt runtime
- Separates thread-scoped Checkpoints from cross-thread Store data and requires a persistent backend for restart recovery
- Resumes through deterministic replay, so non-deterministic work and external effects require tasks, idempotency, and reconciliation
- Supports cooperative graceful drain in LangGraph 1.2 without treating drain as hard task cancellation
- Leaves authorization, tenant isolation, evaluation, and deployment controls to the application stack
Common Use Cases
- Building long-running workflows with conditional branches, cycles, retries, and explicit terminal states
- Pausing a run for externally authorized review and resuming it with a validated Interrupt payload
- Persisting thread state for failure recovery while storing user or application memory across threads separately
- Composing subgraphs or specialist Agents behind typed input, output, State, and permission boundaries
- Replaying or forking recorded trajectories for debugging and evaluation without treating replay as real-world rollback
Example
Loading code...Frequently Asked Questions
Is LangGraph 1.x stable?
LangGraph 1.0 stabilized the core graph APIs and execution model. Later 1.x capabilities can still impose minimum versions; for example, cooperative graceful shutdown requires LangGraph 1.2. Pin the exact package version and test persistence, replay, interrupts, and shutdown behavior before release.
Does LangGraph require an LLM or a Multi-Agent system?
No. A LangGraph Node can run deterministic code, call an API, invoke a model, execute a Tool, or compose another graph. A single workflow can mix these steps, and the Functional API can use ordinary Python control flow. Use multiple Agents only when specialization provides measured value; LangGraph itself does not make a system autonomous or reliable.
What is the difference between a LangGraph Checkpointer and Store?
A Checkpointer records State snapshots for one execution thread and enables conversation continuity, Interrupts, replay, and recovery. A Store holds application-defined data across threads, such as user preferences or shared knowledge. InMemorySaver is process-local and unsuitable for restart recovery; production systems also need retention, access control, encryption, and deletion policies.
Does LangGraph checkpointing prevent duplicate side effects?
No. Resume and interrupt handling can restart a node or entrypoint, and a task that failed after an external write may run again. Put non-deterministic work and side effects in explicit tasks or nodes, use stable operation and idempotency keys, reconcile unknown outcomes, and authorize writes outside the model.
Does LangGraph graceful drain cancel running work?
No. In LangGraph 1.2, a drain request is cooperative and takes effect between supersteps. The current node or retry loop can finish first. Pair drain with a shutdown deadline and downstream cancellation policy, then resume the saved checkpoint using the same thread identity.