My Mega Memory

AI Chat Session Importer and Viewer

Import your AI chat sessions into a local database, then browse and search them through a web interface or via MCP tools.

Supported Providers

Provider	Session Storage Location
Claude Code	~/.claude/projects/
OpenCode	~/.local/share/opencode/storage/
Codex (IntelliJ)	~/.cache/JetBrains/*/aia/codex/sessions/
Codex (CLI)	~/.codex/sessions/
Amp	~/.local/share/amp/threads/
Junie	~/.junie/sessions/
Kilocode	~/.kilocode/cli/
Gemini	~/.gemini/tmp/{project}/chats/
Droid (Factory.ai)	~/.factory/sessions/{project/

Tech Stack

Component	Technology	Purpose
Database	SQLite (better-sqlite3)	Local storage for sessions and messages
Search	SQLite FTS5 (trigram tokenizer)	Full-text search with BM25 ranking
Backend	Node.js + TypeScript + Express 5	Server and CLI application
Templates	EJS + express-ejs-layouts	Server-side rendering
Styling	Tailwind CSS v4 (CDN)	Utility-first CSS framework

Quick Start

# Install dependencies
npm install

# Import your chat sessions
npm run import

# Start the web viewer
npm run dev

npx

npx my-mega-memory import
npx my-mega-memory serve

Then open http://localhost:3000 to browse your sessions.

Create Database

Create fresh databases, dropping all existing data if present:

# npm
npm run create-database

# npx
npx my-mega-memory create-database

Remote Push

You can push sessions from any machine to a running Mega Memory server via API, instead of importing locally.

# npm
npm run push -- --url http://your-server:3000

# npx
npx my-mega-memory push --url http://your-server:3000

This scans all providers locally and sends each session to the server's POST /api/import/session endpoint. The --url flag defaults to http://localhost:3000.

API

Method	Endpoint	Description
POST	/api/import/session	Import a single session. Expects a JSON body with session, provider, projectPath, projectName, created, and updated fields.
POST	/api/mcp	MCP endpoint (Streamable HTTP) with tools: search_sessions, get_session.
GET	/api/mcp	Streamable HTTP session channel (used by MCP clients).
DELETE	/api/mcp	Close a Streamable HTTP MCP session.

The MCP endpoint uses Streamable HTTP over JSON-RPC.

Screenshots