#6666th of 149 in Skills, plugins and rules
memu
Shared memory and skill store for coding agents across sessions and devices
- Stars
- 14.5k
- License
- custom license
- Last commit
- Oct 2026
- Last release
- Mar 2026
- Language
- Python
12,722 on 25 Mar 202614,544 on 14 Oct 2026
From GitHub's daily star counts.
Checked on GitHub.
Overview
memU runs as a sidecar to desktop coding agents (Codex, Claude Code, Cursor, OpenClaw, Hermes, WorkBuddy, Cola, pi). A scheduled task reads session logs, the agent distills them into Markdown memory and skills, and memU embeds and stores them. A standing instruction in the host's instruction file makes the agent run a retrieve command before answering. It can use MemU Cloud or a local SQLite or Postgres store.
Who it is for: Developers using several coding agents who want persistent shared memory
Strengths
- Per-host adapters for Codex, Claude Code, Cursor, OpenClaw, Hermes and others, plus generic memu-agent detect
- Local storage options: SQLite for a single writer, Postgres with pgvector for concurrent access
- Core service makes no LLM calls; the agent writes the skill Markdown, which stays readable
- Embedding providers: OpenAI, Jina, Voyage, Doubao, OpenRouter
Weaknesses
- Chat-only modes of ChatGPT and Claude are not supported for memorize or retrieve
- Self-hosted mode needs an embedding provider API key
- Setup is done by an agent following a SKILL.md, which can fail on some models
- SQLite backend uses brute-force cosine search and allows a single writer
What it needs
- no GPU
- Needs Python 3.11+, Embedding provider API key (self-hosted), SQLite or Postgres with pgvector, MemU Cloud (optional)
- Models: OpenAI embeddings, Jina, Voyage, Doubao, OpenRouter
Also in Skills, plugins and rules
See all 149| Rank | Project | Score |
|---|---|---|
| 1 | cavemanCuts token use in coding agents with terse replies and compressed inputs | 95 out of 100 |
| 2 | skillsAgent skills for planning, TDD, debugging and code review | 94 out of 100 |
| 3 | eccAgents, skills, hooks and rules bundle for Claude Code, Codex and other harnesses | 94 out of 100 |
| #4 | impeccableDesign skill with 24 commands and a detector for AI-generated frontends | 94 out of 100 |
| #5 | ponytailSkill prompt that makes coding agents write less code and add tests | 92 out of 100 |