OpenClaw Cost Optimization Prompts: Cut Token Waste Without Killing Output Quality

Token costs usually spike because teams over-send context and over-request verbose output. Here is the lean operating model.

1) Default to concise mode

Set default response policy to concise and only request deep output when needed.

2) Compress state after long runs

After multi-step execution, ask the agent to produce a compact state snapshot in bullet form.

3) Route by task complexity

• Low-stakes formatting/summarization → cheaper model
• High-stakes architecture/tradeoffs → stronger model

4) Enforce output format constraints

Use fixed sections and max bullet counts. This controls token growth and improves readability.

5) Ask one clarifying question

If scope is ambiguous, ask one targeted question first instead of generating long assumptions that are likely to be thrown away.