#is this token usage normal?
2 messages · Page 1 of 1 (latest)
Short answer: yes, that can be normal in OpenClaw.
Your one-line message is not the whole input. OpenClaw also sends:
- system prompt
- tool schemas
- workspace files (
AGENTS.md,SOUL.md,TOOLS.md, etc.) - session history
- any prior tool results
So 32774 in / 117 out is high for plain chat, but pretty normal for an agentic setup.
Best way to inspect it:
/context detail→ shows what’s eating tokens/usage full→ shows input/output/cacheRead/cacheWrite/compact→ reduces long-thread context
Biggest usual culprits: