Local-first, agent-agnostic memory layer for AI agents. OMem runs quietly on your machine and turns your real work — email, calendar, meeting notes, and files (Office formats, PDFs, images, HTML, Markdown) — into a plain-Markdown wiki that any AI agent can query. Capable agents and your real work context, finally connected.
📚 Documentation: seacen.github.io/omem — concepts, getting started, how-to guides, reference, FAQ (English + 简体中文).
Available on macOS today; Windows is coming. Free for personal use.
Your AI agent is clever, but every session starts blank — it can't see the work going on around you. The genuinely capable agents that arrived this year (Claude Code, Codex, OpenClaw, Hermes) can do real work, but they can't see your inbox, your decks, your meeting notes. OMem is the layer that connects the two.
What makes it different — no other tool sits at this intersection:
- Local-first, zero IT. Reads what your OS already has access to — no vendor cloud, no Graph API consent, no IT ticket.
- The wiki is yours. Plain Markdown in a folder you choose. Read it, grep it, edit it, version it, open it in Obsidian. Nothing locks you in.
- Agent-agnostic. One CLI; a thin skill / MCP wrapper over it. Switch agents next year — your memory persists.
- Search that's actually good. Keyword search by default; enable
qmdfor multi-path semantic + cross-language retrieval. Chinese is a first-class citizen, not an afterthought.
curl -fsSL https://github.com/seacen/omem/releases/latest/download/install.sh | sh
omem setup # guided wizard, ~5 minThen follow the docs — they walk you through setup, your first query, and connecting your agent (Claude Code, Codex, Cursor, …), step by step:
This repo also carries the agent plugin bundle (the OMem skill + manifests) that
the docs install for you; the data layer is the omem CLI binary.
PolyForm Noncommercial 1.0.0 — free for personal / academic / non-profit use; commercial use requires a license.
For commercial licensing, partnership, or any other conversation, get in touch — see About the author or email xichangzhao@outlook.com.
- Documentation: https://seacen.github.io/omem/
- Issues: https://github.com/seacen/omem/issues