Free tools Windows power users keep installed
One-click scans. No signup required.
LangGraph persistence is configured by compiling a graph with a checkpointer and invoking it with a thread_id. The checkpointer saves graph-state snapshots for that thread, enabling continuity, interruption and resumption, and state inspection. For application data that should be shared across threads, configure a separate store. The right checkpointer depends on whether you need a quick in-memory example, lightweight SQLite storage, or durable production persistence.
Checkpointer or store: which kind of persistence do you need?
LangGraph uses two complementary persistence mechanisms. A checkpointer saves state as a graph runs and groups those snapshots into a thread. A store holds application-defined information outside any one thread, making it available across threads. For example, a conversation can use its thread’s checkpointed state for continuity while reading or updating a user’s preferences in a store.
As an Amazon Associate I earn from qualifying purchases.
Checkpointers support continuing conversations, resuming interrupted runs, recovering from failures, inspecting state, and time-travel workflows. Stores are for information your application chooses to retain and share; configure and read or write them explicitly in graph nodes or application code. See the LangGraph persistence guide for the distinction and configuration concepts.
The Tool Desk
Outbyte PC Repair FREEClear out junk files and repair common Windows errorsFree Scan →Outbyte Driver Updater FREEFix the driver behind crashes, sound loss and screen glitchesFind Drivers →What does a checkpoint contain?
A checkpoint is more than a chat transcript. The Python checkpoint reference describes a snapshot of graph state that includes channel values, channel versions, and version tracking for nodes. A thread is a sequence of checkpoints: thread_id identifies the sequence, and checkpoint_id can select a particular snapshot when an operation needs to target one. The LangGraph Python checkpoint reference also documents pending writes: if one node fails after other nodes have completed successfully, those successful writes can be retained so resumption does not necessarily have to rerun all completed work.
Choose a checkpointer for your deployment
Pick an implementation based on durability, concurrency, execution style, operational ownership, and where the graph runs. Official documentation describes intended uses, not universal performance rankings or benchmark results.
| Option | Documented fit | Important trade-off |
|---|---|---|
InMemorySaver / MemorySaver |
Debugging, testing, and the simplest demonstration | State lives in process memory and is lost when the process restarts; it is not durable persistence. |
SqliteSaver |
Lightweight synchronous use, demos, and small projects | The synchronous saver reference says it does not scale to multiple threads. An async SQLite implementation is also available, but its package documentation does not recommend async SQLite for production. |
PostgresSaver / AsyncPostgresSaver |
Durable production workloads and long-running workflows | Requires PostgreSQL connectivity and saver setup. Choose the synchronous or asynchronous integration to suit the application. |
| Agent Server persistence | Managed deployments in which the server handles persistence infrastructure | Backend options depend on the deployment. PostgreSQL is used for server resources and is the default checkpoint backend; supported deployments can use MongoDB as an alternative checkpoint backend, while PostgreSQL remains required for other server resources. |
The SQLite suitability and deserialization guidance is in the SQLite package reference. Agent Server’s persistence arrangement is a managed-deployment detail, not a requirement for standalone LangGraph graphs; see the LangSmith data-plane documentation.
Configure checkpointing in a graph
- Select and initialize a saver. Choose an implementation for your environment and connect it to its backing store. Run any required initialization before compiling; the PostgreSQL example in the checkpoint reference calls
setup(). - Compile the graph with the checkpointer. In Python, pass it as
checkpointer=...when compiling. JavaScript uses the corresponding checkpointer option for graph compilation. - Invoke with a thread identifier. Pass a configuration containing
{"configurable": {"thread_id": "..."}}when invoking the graph. The identifier associates execution with a persisted thread. - Reuse the identifier to continue. Use the same
thread_idwhen continuing that thread; use a different one to create a separate thread. If you need a particular checkpoint rather than the thread’s applicable state, use acheckpoint_idwhere supported by the operation. - Add a store only for cross-thread data. Configure a store separately and have application code or graph nodes explicitly read and write the items that should be shared across threads.
The official quickstart uses an in-memory saver to demonstrate the interface. That makes it useful for learning and tests, but it does not preserve state through a process restart. Follow the implementation-specific setup instructions in the checkpoint reference before using a persistent backend.
Plan retention, identifiers, and graph boundaries
Set a checkpoint retention policy
Checkpoints accumulate over long conversations and can increase latency and storage costs. The persistence guide recommends periodically pruning old checkpoints or setting a retention policy. It does not establish a universal retention period, so choose one based on the application’s recovery, inspection, and storage requirements.
Keep PostgreSQL thread IDs within the documented limit
The persistence guide notes that PostgresSaver stores thread_id in a limited-length column and recommends keeping IDs below 255 characters. A UUID or hash is an option when application identifiers might exceed that length.
Decide how subgraphs share state
A subgraph may use its own checkpoint namespace, which affects what the parent graph can see. If information must cross graph boundaries, the guide suggests using a store or configuring the subgraph to write to the parent checkpoint. Choose deliberately according to whether the information is thread state or shared application data; the persistence guide’s troubleshooting section covers subgraph persistence.
Restrict SQLite checkpoint deserialization where needed
The SQLite package reference documents setting LANGGRAPH_STRICT_MSGPACK=true or passing an explicit allowed_msgpack_modules list to restrict checkpoint deserialization to known-safe types if the database is compromised. Confirm the setting and API against the version of the package installed in your environment.
Outdated Drivers Are Slowing You Down
One free scan finds every outdated or missing driver and matches the right update for your exact hardware.Free scan · exact hardware matchWindows Errors? Fix Them Before They Spread
Repair common Windows errors and clear accumulated junk for a smoother, more stable PC - no reinstall needed.Free scan · no reinstallQuick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
