memex MCP
Persistent memory for coding agents: insights are stored as markdown cards and recalled at the start of the next session.
- 144 ★ on the repository
- Last commit
- MIT
Extended access
Why this level
- The agent writes cards to `~/.memex/cards/` itself, and git sync can push them to a remote repository.
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add mcp/iamtouchskyer-memexDetects the agents on your machine, shows the access level and pins the version.
Other ways to install
Access and checks
How we reviewExtended accessThe tool runs code or makes network calls and only reads external data.
- The agent writes cards to `~/.memex/cards/` itself, and git sync can push them to a remote repository.
This is third-party code. Review the repository files before installing.
What it does
memex saves what an agent learned during a task as atomic cards with bidirectional `[[wiki-style]]` links. In the next session the agent first recalls relevant cards and continues instead of starting from scratch. There is no vector database or embeddings, only markdown files a human can read too. Claude Code connects the memory through a plugin with hooks and skills; Cursor, Codex, Windsurf and VS Code Copilot use an MCP server with 12 tools. All clients write to one `~/.memex/cards/` directory, so a card from Claude Code is available in Cursor. `memex sync` moves cards between machines through git, and `memex serve` shows a timeline and link graph at `localhost:3939`.
Who it is for. People who use several agents at once and are tired of re-explaining the project every time.
Good fit when
- Yesterday the agent figured out an odd build and today it asks how it works again.
- You switch between Claude Code and Cursor and want one shared knowledge base.
- Project notes must live in plain files you can read and edit by hand.
- You need to sync memory between a laptop and a work machine through a private repository.
Not a fit when
- You need semantic search over large documents with embeddings.
- Notes contain secrets or third-party data that must not go into a synced repository.
- One agent and one-off tasks where memory between sessions adds nothing.
Example request
Recall what we learned about this project's build and save today's fix as a card linked to the earlier ones.MCP
- Transport
- stdio
- Authentication
- not required
Limitations
Every client except the Claude Code plugin and the VS Code extension needs Node.js 18 or newer and a global `@touchskyer/memex` install. Creating the sync repository automatically requires GitHub CLI (`gh auth login`); otherwise pass your own repository URL, GitLab included. After sync is configured, `memex serve` opens the memra.vercel.app web interface and sends queries there; `--local` keeps everything on the machine. The agent writes the cards itself, so review the content.
How to disable. In Claude Code run `/plugin uninstall memex`. For MCP remove the server from the client configuration and run `npm uninstall -g @touchskyer/memex`. Cards stay in `~/.memex/cards/`; delete that directory separately.
FAQ
How does Claude Code differ from other clients?
Claude Code adds automatic recall through a session-start hook and slash commands. Cursor, Codex and Windsurf use the same cards through 12 MCP tools.
How do I move memory to another computer?
Run `memex sync --init`: it creates a private `memex-cards` repository through GitHub CLI. Pass your own with `memex sync --init <repo-url>`. Then `memex sync on` enables auto-sync.
How can I browse the cards?
Run `memex serve`: a timeline and link graph open at `localhost:3939`. The `--local` flag prevents the external web interface from opening.
Is a vector database needed?
No. Search works over markdown files and card links; embeddings are not used.
Which Node.js version is needed?
Node.js 18 or newer for every client except the Claude Code plugin and VS Code extension, which bundle everything.
Discussion
No questions or reviews yet.
Related
MCP serversLong-term agent memory in local Markdown files with a link graph, semantic search and MCP access
A long-term memory plugin that keeps decisions, preferences and task state across sessions and injects only what a query needs
Long-term memory for Claude Code, Cursor and Codex: the agent saves decisions and finds them again in later sessions
A local workbench for Claude Code, Codex, Pi and OpenCode: memory changes wait for your review, then an audit follows each turn
Long-term memory for agents: it distills knowledge from session history and brings it back in new tasks, one store shared across Claude Code, Codex, Cursor and
Agent memory as a graph of linked facts: decisions and errors outlive the session, with no vector database or paid API calls
Comparison with similar tools
Cards that share the main search query, and how they differ.
| Tool | Official | Authentication | Repository stars | Main difference |
|---|---|---|---|---|
| memex (this card) | no | not required | 144 | |
| Basic Memory | no | not required | 4,495 | markdown notes with a different storage format; here cards with bidirectional links and one directory for all clients |
| Claude-Mem | no | 113,534 | memory for Claude Code; here the same memory is available in Cursor, Codex and Windsurf |
README in short
All clients read and write one card directory. The `memex organize` command reports orphans, hubs and conflicts in the link network.