#121121st of 149 in Skills, plugins and rules
agent-memory
Markdown-file long-term memory for coding agents, with local SQLite search
- Stars
- 3.6k
- License
- MIT
- Last commit
- Oct 2026
- Language
- Python
Missed the latest GitHub check; these facts may be older.
Overview
agent-memory stores agent memories as markdown files in one store, with a rebuildable SQLite FTS5 index (BM25, optional vector plugin fused by RRF) for local ranked recall. Hooks installed for Claude Code, Codex CLI and Muse Code distil memories at conversation boundaries, and a sleep-time pass consolidates them, with deletions only as proposals. It exposes a `mem` CLI and a `mem-mcp` stdio MCP server.
Who it is for: Developers wanting persistent, inspectable memory for Claude Code or Codex
Strengths
- Files are the source of truth; `rm -rf .index/ && mem rebuild` loses nothing, enforced by a test
- No LLM client in the library; reasoning borrows the host agent's CLI, so no API key required
- One store shared across Claude Code, Codex CLI and Muse Code; any agent that runs shell commands works
- Recall returns paths and abstracts, so the agent opens files only as deep as needed
Weaknesses
- No PyPI release; install is from a git checkout with uv sync
- Benchmark is on a bounded LongMemEval-S haystack, so scores are not comparable to published ones
- `context`, `sleep` and the proposal ledger have no MCP tool yet; CLI only
- Muse Code setup needs OpenRouter credentials and a separate pproxy dependency
What it needs
- no GPU
- Needs Python 3.12+, uv, SQLite FTS5, pproxy (Muse Code only), gcloud CLI (optional, Vertex AI reasoner)
- Models: Claude Code, Codex CLI, Muse Code, Hermes, Vertex AI endpoint
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 |