OpenParable
Skip to content
← Back to skills

Memu

Apps & Automation
memu

Memu is an Agent Skill for apps & automation published in the NevaMind-AI/memU repository on GitHub: a SKILL.md instruction file an AI agent can load to do this task.

Give this agent persistent memory via memU. Use to memorize (save files/folders so context survives this session) when the user asks to remember or sync something, or after completing work worth keeping; and to retrieve (search memory from earlier sessions) when the user asks what is known about a person/project/topic or references context not in this conversation.

View on GitHub ↗
QUICK START

How to use this skill

Bring this guide into your coding agent with a prompt tailored to the tool you use.

  1. Open your project in Codex.
  2. Copy and paste the prompt below.
  3. Review files and risks before approving any action.
Prompt to paste
Inspect the Agent Skill at https://github.com/NevaMind-AI/memU/blob/HEAD/.claude/skills/memu/SKILL.md and all supporting files. Report provenance, permissions, scripts, dependencies, and security risks. Propose an exact installation into .agents/skills/memu, but wait for my approval before writing files or running scripts.

Copying this prompt does not install or run the skill. Review third-party files before use.

memU: memorize and retrieve persistent memory

memU compiles sources into a persistent local store (./data/memu.sqlite3 + a browsable ./data/memory/ markdown tree, relative to CWD). What one session memorizes, the next can retrieve — always run from the project root so every command hits the same store.

Use only the two commands below. Do not use the legacy memu memorize / memu retrieve commands.

Both directions need an API key: make sure OPENAI_API_KEY (or MEMU_LLM_PROVIDER + its matching key) is set, and tell the user if it is missing.

Locate the CLI

Use the first available runner:

  1. memu (installed via pip install memu-py)
  2. uvx --from memu-py memu
  3. npx memu-cli

Memorize

memu memorize-workspace <folder>
  • Incremental and safe to re-run: diffs against <folder>/.memu_manifest.json, so only added/modified files are processed and memory from deleted files is removed.
  • Top-level directory decides the treatment: chat/ → memory topics, agent/ → skills, everything else → indexed workspace context (modality inferred per file from its extension).
  • To memorize a single file, place (or copy) it into the workspace folder and sync — there is no separate single-file path.

Report the printed diff (added/modified/deleted) to the user; pass --json if you need to parse the result.

Retrieve

memu retrieve-workspace "<query>"

Single-shot embedding search, no LLM calls. Returns JSON in three layers:

  • segments — the matched slices, ranked by similarity (check score)
  • files — the memory/skill documents they belong to (usually what you want)
  • resources — matching raw sources, when summaries are not enough

If a query misses, retry with different phrasing or more specific terms. Low scores across the board usually mean nothing relevant is stored — say so rather than stretching weak matches. You can also read ./data/memory/MEMORY.md / SKILL.md / INDEX.md directly for a browsable overview.