Memex

AI CLI Session History Manager - On-demand retrieval, precise recall

One Memory. All CLIs. Never Compacted. Exact Search.

What is Memex?

Memex adds long-term memory to AI coding assistants through on-demand search. Instead of losing your valuable conversations, you can search and retrieve precise context whenever you need it.

Supported Tools

✅ Claude Code
✅ Codex CLI
✅ OpenCode
✅ Gemini CLI

Features

On-demand search - You control when to search; automatic injection is opt-in
Original preservation - Raw messages always kept; summaries are optional layers
Multi-CLI support - Claude Code, Codex, OpenCode, Gemini in one database
Powerful search - Full-text (FTS5) + semantic vectors + hybrid ranking
MCP integration - Search directly from your AI CLI
REST API - Integrate into any workflow
Local storage - All data stays on your machine

Why Memex? Read the full explanation in Why Memex?

Quick Start

Homebrew (macOS / Linux)

brew install vimo-ai/tap/memex

# Verify server is running
curl http://localhost:10013/health
# OK

# Zero-dependency CLI, reads local sessions directly
brew install vimo-ai/tap/memex-lite

memex search "anything you want"
memex list -n 10

Docker

docker run -d -p 10013:10013 \
  -v ~/.vimo:/data \
  -v ~/.claude/projects:/claude:ro \
  -v ~/.codex:/codex:ro \                              # 可选: Codex
  -v ~/.local/share/opencode:/opencode:ro \            # 可选: OpenCode
  -v ~/.gemini/tmp:/gemini:ro \                        # 可选: Gemini
  -e OLLAMA_HOST=http://host.docker.internal:11434 \   # 可选: 本机 Ollama (Docker Desktop)
  ghcr.io/vimo-ai/memex:latest

docker run -d -p 10013:10013 `
  -v "$env:USERPROFILE\.vimo:/data" `
  -v "$env:USERPROFILE\.claude\projects:/claude:ro" `
  -v "$env:USERPROFILE\.codex:/codex:ro" `             # 可选: Codex
  -v "$env:LOCALAPPDATA\opencode:/opencode:ro" `       # 可选: OpenCode
  -v "$env:USERPROFILE\.gemini\tmp:/gemini:ro" `       # 可选: Gemini
  -e OLLAMA_HOST=http://host.docker.internal:11434 `   # 可选: 本机 Ollama (Docker Desktop)
  ghcr.io/vimo-ai/memex:latest

Configure MCP (Full / Docker)

claude mcp add memex -- npx -y mcp-remote http://localhost:10013/api/mcp

codex mcp add memex -- npx -y mcp-remote http://localhost:10013/api/mcp

gemini mcp add --transport http memex http://localhost:10013/api/mcp

# Edit ~/.config/opencode/opencode.json
{
  "mcp": {
    "memex": {
      "type": "remote",
      "url": "http://localhost:10013/api/mcp"
    }
  }
}

Then search in your AI CLI:

use memex search "anything you want"

Documentation

Installation - Docker, build from source, troubleshooting
Configuration - Environment variables and options
API Reference - REST API endpoints
MCP Tools - Claude Code integration
How it Works - Data flow, architecture, and internals
Advanced - Claude Code Hooks and more