memex

Overview

Memex provides persistent memory for AI coding agents using a Zettelkasten-based approach. It allows agents to remember insights across sessions by saving knowledge cards with bidirectional links. The project offers MCP server integration that exposes 10 tools for various AI platforms including Claude Code, VS Code, Cursor, and others. Instead of using vector databases, it stores knowledge as plain markdown files in ~/.memex/cards/, making it human-readable and easily portable. The system supports cross-platform sharing and git synchronization to keep memories consistent across devices and tools.

Try asking AI

After installing, here are 5 things you can ask your AI assistant:

• memex_search

  Search for cards or list all cards

  Try:Search for cards about PythonTry:List all my cards
• memex_read

  Read a specific card by slug

  Try:Read my card about async programmingTry:Show the contents of python-tips
• memex_write

  Write a new card with content from stdin

  Try:Create a new card about React hooksTry:Write a card about database optimization
• memex_links

  Show link graph statistics for a card

  Try:Show links for my machine learning cardTry:Display link graph stats for python-basics
• memex_archive

  Archive a card

  Try:Archive my old project notesTry:Move this completed task to archive
• memex_serve

  Start visual timeline UI to browse memory

  Try:Open the memex timeline viewTry:Show my knowledge graph
• memex_sync

  Sync cards via git

  Try:Sync my cards to GitHubTry:Enable automatic sync after writing
• memex_mcp

  Start MCP server (stdio)

  Try:Start memex MCP serverTry:Initialize memex for AI integration

Note: Inferred tool names from CLI reference section. The documentation mentions '10 MCP tools' but doesn't explicitly list them all. Tool names are based on the CLI commands that would be exposed via MCP.

Comparable tools

Installation

{
  "mcpServers": {
    "memex": {
      "command": "memex",
      "args": ["mcp"]
    }
  }
}

FAQ

How is Memex different from other memory systems?

Memex uses Zettelkasten with bidirectional links instead of vector databases or embeddings. It stores knowledge as readable markdown files rather than opaque embeddings, making it human-readable and easily portable.

Does Memex support multiple AI coding platforms?

Yes, it supports Claude Code, VS Code/Copilot, Cursor, Codex, Windsurf, and any MCP client. All platforms share the same memory store through ~/.memex/cards/ directory.

Compare memex with