Stats & diagnostics

​

What this means in practice

• Monitor total memory count and archive status
• Understand memory distribution across stores (sensory, episodic, semantic, procedural, prospective)
• Check recent capture activity (captures in the last 24 hours)
• Identify the strongest memories by strength score
• Assess overall health and usage patterns

​

Parameters

​

Returns

{
  "totalMemories": 342,
  "archivedCount": 28,
  "avgStrength": 0.756,
  "recentCaptures": 12,
  "byStore": {
    "sensory": 89,
    "episodic": 156,
    "semantic": 54,
    "procedural": 32,
    "prospective": 11
  },
  "strongest": [
    {
      "title": "Chose Convex over Supabase for backend",
      "store": "semantic",
      "strength": 0.95,
      "category": "decision",
      "id": "mem_xyz123"
    },
    {
      "title": "Deploy via Railway + git push",
      "store": "procedural",
      "strength": 0.91,
      "category": "workflow",
      "id": "mem_abc456"
    }
  ]
}

## Memory Crystal Stats
- Total memories: 342
- Archived memories: 28
- Average strength: 0.756
- Captures last 24h: 12

## Store counts
- sensory: 89
- episodic: 156
- semantic: 54
- procedural: 32
- prospective: 11

## Strongest
1. Chose Convex over Supabase for backend (semantic) — 0.95
2. Deploy via Railway + git push (procedural) — 0.91

​

Examples

{}

{
  "channel": "dev-team"
}

​

Interpreting the results

• totalMemories — cumulative memories ever created (including archived)
• archivedCount — memories marked as archived; not recalled by default
• avgStrength — mean strength across all active memories; indicates overall salience
• recentCaptures — automatic captures in the last 24 hours from crystal_capture hook
• byStore — distribution across the five memory types
• strongest — top-N memories by strength; useful for identifying what the system prioritizes

​

When to use vs alternatives

​

Error handling

​

Source of truth

• mcp-server/src/tools/stats.ts — tool definition
• convex/crystal/stats.ts — backend calculation logic
• apps/web/app/(dashboard)/ — stats visualization in the dashboard