Overview
How might we design cost transparency so people can trust and act on AI output?
When to use
- Ideal for applications with usage-based pricing, credit systems, and platforms where cost awareness helps users manage spending and optimize usage.
When to skip
- Flat unlimited consumer plans where showing micro-costs adds anxiety with no decision value.
- Internal demos on company-paid keys where cost UI distracts from the task.
- When estimates are so wrong that displaying them destroys trust.
Rules
Hiding cost until after a large agent run finishes.
Inconsistent units (credits vs dollars) across model picker and usage page.
“Free” labeling for actions that still burn shared team quotas.
No warning when a tool call will spend an order of magnitude more than a normal reply.
Evidence
| Product | Implementation |
|---|---|
| OpenAI / Anthropic consoles | Token and dollar usage visible per request and in billing dashboards. |
| Cursor | Surfaces premium model and agent usage against plan limits. |
| Perplexity / Claude Pro | Makes plan limits and upgraded modes legible at choice time. |
| Cloud LLM gateways | Per-team budgets and cost estimates before high-spend agent jobs. |



