Compute Budget
See the cost before you spend it
Before an agent runs, surface its expected token, time, and dollar cost. Pick a mode (Quick / Standard / Deep) that scales the budget. Mid-flight, a meter shows what you are actually spending — and pauses the run when it crosses the cap so you can decide whether to keep going.
Framing
The problem
Agent runs can spend tokens, time, and money invisibly — users only learn the cost after the bill arrives.
The pattern
Show pre-flight cost estimates with mode selection and a hard cap; meter the live spend and pause at threshold.
Why chat breaks here
Chat hides compute as a black box — there is no surface to set a ceiling, see live spend, or decide mid-run whether the answer is worth more.
Risks
Cost meters can become anxiety theatre if they are imprecise, distracting, or punish exploration.
Avoid when
The task is cheap enough end-to-end that a budget interface costs more attention than it saves money.
Use when
Agent runs spend tokens, time, or money in ways users cannot see or limit at the start.
DOPE evaluation
- Directability
- Raise the cap, switch mode, or stop and pocket partial output any time
- Observability
- Live meter shows tokens, time, and dollars consumed against the budget
- Predictability
- Mode and cap are explicit upfront — no runaway spend hidden in a long run
- Explainability
- Estimate breakdown shows which steps and tools drive the cost
In the wild
- Replit Effort-Based Pricing (Replit) — Per-checkpoint cost shown as work happens (~$0.25 per simple checkpoint, more for complex). Effort/credit display directly in the agent chat surface. Strongest end-user fit (July 2025).
- Devin ACUs (Cognition) — ACU meter (1 ACU ≈ 15 min, ~$2.25/ACU on Core, $2.00 on Team). Per-task ACU consumption shown in dashboard/timeline. Pay-as-you-go added April 2025.
- Claude Code /cost (Anthropic) — Per-session token + dollar estimate displayed inline in the terminal via /cost and /usage commands. 5-hour rolling window quota visible.
- Cursor Request Meter (Cursor) — Per-request credit/usage display for premium models in the bottom bar; Cursor 2.x surfaces a "high-cost run" warning before agent calls.
- OpenAI Deep Research modes (OpenAI) — Caveat — Tiered modes (Light/Standard/Deep on Pro) match the pre-flight mode-selection half of the pattern, but the live dollar meter is API-side, not in the chat surface.
FAQ
When should I use the Compute Budget pattern?
Agent runs spend tokens, time, or money in ways users cannot see or limit at the start.
When should I avoid the Compute Budget pattern?
The task is cheap enough end-to-end that a budget interface costs more attention than it saves money.
What problem does Compute Budget solve?
Agent runs can spend tokens, time, and money invisibly — users only learn the cost after the bill arrives.
Why is chat the wrong fit for this?
Chat hides compute as a black box — there is no surface to set a ceiling, see live spend, or decide mid-run whether the answer is worth more.
Related patterns
- Often paired with: Draft Mode — Budget caps the run; Draft Mode caps the model. Together they make exploration cheap and commitment explicit.
- Alternative to: Approval Gates — Stop on cost vs stop on step. Both are pause points — Budget is the economic gate, Approval is the human one.
- Often paired with: Autonomy Dial — Set how much autonomy the agent has, then bound how much it can spend exercising it.