Mnemo

Persistent engineering cognition for AI coding agents. Your agent remembers decisions, understands code relationships, and never asks the same question twice.

Features

• Auto-init — prompts to initialize when you open a workspace without .mnemo/
• Client detection — auto-detects Kiro, Amazon Q, Cursor, Claude Code; asks if none found
• Status bar — shows Mnemo status (Active/Inactive)
• Refresh index — re-scan codebase after major changes
• Doctor — diagnose installation and MCP connectivity

Commands

How It Works

1. Extension activates on workspace open
2. If .mnemo/ doesn't exist → prompts "Initialize project memory?"
3. Auto-detects your AI client (Kiro, Amazon Q, Cursor, Claude Code)
4. If multiple or none detected → shows picker
5. Runs mnemo init --client <choice> which:
   • Builds knowledge graph (classes, methods, calls, communities)
   • Installs lifecycle hooks for your client
   • Configures MCP server
   • Downloads embedding model (one-time, 86MB)

What Your Agent Gets

After initialization, your AI agent has access to 16 MCP tools:

• Memory: recall, remember, forget, decide, search
• Code Intelligence: lookup, graph, impact, search, communities
• Planning: create plans, track progress, mark done
• Safety: security scan, dead code, conventions audit

Requirements

• Python 3.10+
• pip install mnemo-dev (or extension auto-downloads binary)

Supported Clients

Performance

• Init: 3-7 seconds (depending on repo size)
• Search: 2ms (semantic + keyword + graph)
• RAM: ~265 MB
• No external databases or servers