MemoryAtlas

Procedural memory — learns how to do tasks and rewrites the agent's own behavior/prompts.

Storage
Whatever your LangGraph store is (e.g. Postgres)
Retrieval
Retrieves from your configured LangGraph store across semantic, episodic, and procedural memory types.
Self-host
Self-host: trivial
License
MIT
Pricing
Free OSS SDK; managed service · Free + paid
GitHub stars
1,602
Last release
2025-10-27
Last commit
2026-08-05
First catalogued
2026-06-28

Strengths

  • Semantic + episodic + procedural memory types in one SDK
  • Procedural memory via create_prompt_optimizer: rewrites system prompts from episodic feedback
  • Hot-path and background formation modes — no latency hit for pattern extraction
  • Native LangGraph store integration; works with any LangChain/LangGraph stack

Watch out

  • Not a turnkey backend — you provision and wire your own storage (Postgres, InMemory, etc.)
  • LangGraph coupling: procedural optimizer and background manager assume LangGraph agent loop
  • Latest PyPI release is v0.0.30 (Oct 2025); repo has active commits through Jun 2026 but no new package release in ~8 months — pin to v0.0.30 or pin the git sha for reproducibility

Best for

  • Teams already on LangChain/LangGraph who need agents that improve from feedback

How it integrates

Benchmark results

No sourced results yet.

Sources

Last verified 2026-08-01 · updated by refresh-framework-cards