---
title: "LangMem and past.dev: SDK memory vs memory service"
description: "LangMem manages long-term memories over the LangGraph store. Compare its responsibilities, storage model and operation with the past.dev API."
canonical: https://past.dev/compare/langmem
last-updated: 2026-08-31
---
# LangMem and past.dev

Source: https://past.dev/compare/langmem

LangMem extracts and manages long-term memories for agents using a LangGraph store that the application configures and operates. past.dev provides extraction, entity resolution and temporal storage as an API service. Its data model includes validity windows, supersession and explicit recall status.

## What LangMem provides

[LangMem](https://langchain-ai.github.io/langmem/) provides utilities for extracting memories from conversations and storing them in a [LangGraph store](https://langchain-ai.github.io/langgraph/concepts/persistence/). Entries use namespaces and can include embeddings for semantic lookup. It provides cross-thread memory within the LangGraph ecosystem.

## What remains your job with an SDK

- Operating the store: the backing persistence, its scaling and its versions are configuration you own.
- The time model: store entries are keys with values; writing a key replaces it. Validity windows, supersession links and as-of queries are not part of the model.
- Identity across sources: entries are namespaced by the keys you choose; joining the same person across email, chat and tickets is application logic.
- Evidence status: empty retrieval returns an empty list. Application logic must distinguish insufficient evidence from a retrieval miss.

## When to use each product

- **Choose LangMem** for LangChain-native personalization where current-state preferences per user are the requirement and you are happy operating the store.
- **Choose past.dev** when changing facts require previous values, dates, sources, and explicit `Conflicted` or `NoKnownSupport` status. The [LangGraph integration](/integrations/langgraph) uses checkpointers and stores for graph state, and the API for temporal facts.

## Evaluation test

Test a fact that changes, such as a user's employer. Ask for the current value and the value before the change. Supporting both requires [supersession](/glossary/fact-supersession) and [point-in-time recall](/glossary/point-in-time-recall). The [quickstart](/docs/memory-api/quickstart) demonstrates both. The [benchmarks page](/benchmarks) documents update tests.

## Frequently asked questions

### Can LangMem and past.dev be used together?

Yes. A common split keeps LangMem for lightweight per-user preferences inside the app and routes durable, changing, shared facts through the API as tools. The LangChain and LangGraph integration pages show the wiring.

## Related

- [Memory for LangChain](https://past.dev/integrations/langchain)
- [Memory for LangGraph](https://past.dev/integrations/langgraph)
- [Point-in-time recall](https://past.dev/glossary/point-in-time-recall)
- [What is agent memory?](https://past.dev/what-is-agent-memory)
- [Quickstart](https://past.dev/docs/memory-api/quickstart)