Back to Skills
// SKILLMemory

Memory Tiering

Multi-tiered memory management that organizes your bot's data into HOT, WARM, and COLD tiers for efficient retrieval and cost optimization.

/memory-tiering

KEY FEATURES

1

Automatic classification into HOT (active), WARM (recent), and COLD (archive) tiers

2

Smart promotion and demotion between tiers based on access patterns

3

Configurable retention policies per tier

4

Semantic search across all memory tiers

5

Memory compression for COLD tier to save token costs

CONFIGURATION EXAMPLE

openclaw.json
"skills": {
  "entries": {
    "memory-tiering": { "enabled": true }
  }
}

EXAMPLE CONVERSATION

U

What did we discuss about project Alpha last week?

B

Retrieving from WARM memory... Last Tuesday you shared the project Alpha requirements: - Budget: €50K for Q2 - Deadline: March 15 - Team: 4 developers You also mentioned wanting to use React for the frontend. Want me to promote this to HOT memory for faster access?

TIPS & BEST PRACTICES

•

Set shorter WARM retention for chatbots with high message volume

•

COLD tier data is compressed — retrieval is slower but more cost-effective

•

Pair with Context Budgeting for maximum token savings

RELATED SKILLS

Context BudgetingAI Optimization

Enable on your bot

Memory Tiering · Memory

DASHBOARD
Back to Skills