memory

Claude Mind

You have access to a persistent memory system powered by Claude Mind. All your observations, discoveries, and learnings are stored in a single .agent-brain/mind.mv2 file.

How to Execute Memory Commands

Use the bundled SDK scripts via Node.js (NOT the CLI). The scripts are at ${CLAUDE_PLUGIN_ROOT}/dist/scripts/.

Search Memories

node "${CLAUDE_PLUGIN_ROOT}/dist/scripts/find.js" "<query>" [limit]

Examples:

• node "${CLAUDE_PLUGIN_ROOT}/dist/scripts/find.js" "authentication" 5
• node "${CLAUDE_PLUGIN_ROOT}/dist/scripts/find.js" "database schema" 10

Ask Questions

node "${CLAUDE_PLUGIN_ROOT}/dist/scripts/ask.js" "<question>"

Examples:

• node "${CLAUDE_PLUGIN_ROOT}/dist/scripts/ask.js" "Why did we choose React?"
• node "${CLAUDE_PLUGIN_ROOT}/dist/scripts/ask.js" "What was the CORS solution?"

View Statistics

node "${CLAUDE_PLUGIN_ROOT}/dist/scripts/stats.js"

View Recent Memories

node "${CLAUDE_PLUGIN_ROOT}/dist/scripts/timeline.js" [count]

Memory Types

Memories are automatically classified into these types:

• discovery - New information discovered
• decision - Important decisions made
• problem - Problems or errors encountered
• solution - Solutions implemented
• pattern - Patterns recognized in code/data
• warning - Warnings or concerns noted
• success - Successful outcomes
• refactor - Code refactoring done
• bugfix - Bugs fixed
• feature - Features added

File Location

Your memory is stored at: .agent-brain/mind.mv2

This file is:

• Portable - Copy it anywhere, share with teammates
• Git-friendly - Commit to version control
• Self-contained - Everything in ONE file
• Searchable - Instant semantic search

Usage Tips

1. Start of session: Recent memories are automatically injected as context
2. During coding: Observations are captured automatically from tool use
3. Searching: Use natural language queries to find relevant past context
4. Sharing: Send the .mind.mv2 file to teammates for instant onboarding