#3030th of 30 in Memory

simplemem

Long-term memory for LLM agents, with text and multimodal backends

Stars
3.9k
License
MIT
Last commit
Jul 2026
Last release
May 2026
Language
Python
Stars, last 29 weeks+813
25 Apr 2026: 3,070 stars2 May 2026: 3,094 stars9 May 2026: 3,113 stars16 May 2026: 3,139 stars23 May 2026: 3,331 stars30 May 2026: 3,397 stars6 Jun 2026: 3,423 stars13 Jun 2026: 3,455 stars20 Jun 2026: 3,474 stars27 Jun 2026: 3,516 stars4 Jul 2026: 3,537 stars11 Jul 2026: 3,589 stars18 Jul 2026: 3,606 stars25 Jul 2026: 3,618 stars1 Aug 2026: 3,631 stars8 Aug 2026: 3,645 stars15 Aug 2026: 3,675 stars22 Aug 2026: 3,682 stars29 Aug 2026: 3,700 stars5 Sept 2026: 3,727 stars12 Sept 2026: 3,732 stars19 Sept 2026: 3,746 stars26 Sept 2026: 3,791 stars3 Oct 2026: 3,810 stars10 Oct 2026: 3,821 stars17 Oct 2026: 3,837 stars24 Oct 2026: 3,851 stars31 Oct 2026: 3,860 stars7 Nov 2026: 3,876 stars14 Nov 2026: 3,883 stars

3,070 on 25 Apr 20263,883 on 14 Nov 2026

From GitHub's daily star counts.

Checked on GitHub.

Overview

SimpleMem is a Python package that compresses dialogue into atomic, timestamped memory units and retrieves them by inferred query intent. One `simplemem` import routes to a text backend or an Omni backend (image, audio, video; FAISS plus BM25), and an EvolveMem step tunes retrieval settings on a dev set. It can run as an MCP server, either cloud-hosted or self-hosted via Docker Compose with a web UI on port 8000.

Who it is for: Developers adding persistent memory to LLM agents or MCP clients

Strengths

  • Auto-routes between text and multimodal backends from the first method called
  • Works with MCP clients such as Claude Desktop, Cursor, LM Studio and Cherry Studio
  • Offline retrieval-config tuning against your own dev questions via simplemem.optimize
  • Self-hostable with docker compose; data persists in ./data (LanceDB)

Weaknesses

  • Needs an OpenAI-compatible API key for memory construction and retrieval
  • Cloud MCP endpoint is the default path; self-hosting needs JWT and encryption keys
  • Full multimodal use requires the Python API; MCP covers text memory only
  • Benchmark gains (LoCoMo F1, 30x token cut) are self-reported from the authors' papers

What it needs

  • Docker + Compose
  • Needs OpenAI-compatible LLM API, LanceDB, FAISS, Docker (optional)
  • Models: OpenAI, Qwen, Azure OpenAI, Ollama, Requesty
  • port 8000

What changed

From the release notes. No flag means the note does not say so, not that the upgrade is safe.

Also in Memory

See all 30
Also in Memory
RankProjectScore
1hindsightAgent memory server with retain, recall and reflect operations61.7k starsMIT95 out of 100
2gbrainMemory store with sources and a knowledge graph for coding agents33.2k starsMIT93 out of 100
3MemPalaceLocal verbatim memory store for AI agents with semantic search and MCP60.6k starsMIT90 out of 100
#4SupermemoryMemory and context engine for AI agents, with user profiles and hybrid search31.7k starsMIT89 out of 100
#5Mem0Long-term memory layer that stores and retrieves user facts for LLM agents Live demo ↗ (opens in a new tab)69.2k starsApache-2.088 out of 100