Persistent memory for AI coding agents. Your agent remembers what it learned across sessions.
Every time your AI agent finishes a task, it saves insights as atomic knowledge cards with [[bidirectional links]]. Next session, it recalls relevant cards before starting work — building on what it already knows instead of starting from scratch.
Session 1: Agent fixes auth bug → saves insight about JWT revocation
Session 2: Agent works on session management → recalls JWT card, builds on prior knowledge
Session 3: Agent organizes card network → detects orphans, rebuilds keyword index
No vector database, no embeddings — just markdown files your agent (and you) can read.
| Platform | Command |
|---|---|
| Any editor | npx add-mcp @touchskyer/memex -- mcp |
| Claude Code | /plugin marketplace add iamtouchskyer/memex then /plugin install memex@memex |
| VS Code / Copilot | Install from MCP Registry or code --add-mcp '{"name":"memex","command":"npx","args":["-y","@touchskyer/memex","mcp"]}' |
| Cursor | One-click install |
| Codex | codex mcp add memex -- npx -y @touchskyer/memex mcp |
| Windsurf / others | Add MCP server: command npx, args ["-y", "@touchskyer/memex", "mcp"] |
Then, in your project directory:
npx @touchskyer/memex initThis adds a memex section to AGENTS.md that teaches your agent when to recall and retro. Works with Cursor, Copilot, Codex, and Windsurf. Claude Code users don't need this — the plugin handles it.
| Platform | How |
|---|---|
| npx users (VS Code, Cursor, Windsurf) | Automatic — npx -y always fetches latest |
| Claude Code | npm update -g @touchskyer/memex (plugin updates from marketplace) |
| Codex / global install | npm update -g @touchskyer/memex |
memex serveOpens a visual timeline of all your cards at localhost:3939.
memex sync --init git@github.com:you/memex-cards.git
memex sync on # enable auto-sync after every write
memex sync # manual sync
memex sync off # disable auto-syncCards are plain markdown — git handles merging and history.
memex search [query] # search cards, or list all
memex read <slug> # read a card
memex write <slug> # write a card (stdin)
memex links [slug] # link graph stats
memex archive <slug> # archive a card
memex serve # visual timeline UI
memex sync # sync via git
memex mcp # start MCP server (stdio)
memex init # add memex section to AGENTS.mdBased on Niklas Luhmann's Zettelkasten method — the system behind 70 books from 90,000 handwritten cards:
- Atomic notes — one idea per card
- Own words — forces understanding (the Feynman method)
- Links in context — "this relates to [[X]] because..." not just tags
- Keyword index — curated entry points to the card network
Cards are stored as markdown in ~/.memex/cards/. Open them in Obsidian, edit with vim, grep from terminal. Your memory is never locked in.
MIT