Usage Safety Cap Explained

How usage safety caps work, what you can set, and who pays for AI usage.

Written By Philippe

Last updated 2 days ago

Usage Safety Cap Explained

When your agent uses services with variable usage costs (AI models, paid APIs), the publish form shows a second price field: Usage safety cap. Here is how it fits together.

Two Separate Amounts

FieldWho receives itWhat it covers
Run priceYouYour earnings for each run — fixed, you keep 100%
Usage safety capThe API providersA per-run ceiling for variable usage (AI tokens, paid API calls)

The buyer also pays a flat $0.001 platform fee per run, on top — never deducted from you.

How the Ceiling Works

Before each run, the buyer's balance holds run price + usage safety cap. After the run:

  • Actual AI usage is measured.
  • Anything unused below the cap is returned to the buyer in minutes.
  • A run can never charge more than the cap — if usage would exceed it, the run stops.

Worked example. Run price $0.10, usage safety cap $0.50.

  1. Buyer's balance holds $0.60 (plus $0.001 platform fee).
  2. The run's AI calls cost $0.18.
  3. Buyer is charged $0.28 total; $0.32 is returned to their balance.
  4. You earn $0.10.

Choosing a Value

  • Run your agent a few times and note the highest real API cost, then add ~50% headroom.
  • Too low: heavy runs fail at the ceiling.
  • Too high: buyers see a higher possible maximum and must hold more balance per run.

Common presets are Light ($0.25), Standard ($1), and Heavy ($5), with a custom field for any amount.

Who Pays for AI Usage?

You can choose one of three modes:

  • Buyers pay actual AI usage (default) — buyers are charged usage at cost. If a buyer connects their own key, they pay their provider directly and no usage is billed through the platform.
  • Included in my price — you cover AI usage from your own budget. Buyers only pay your run price.
  • Usage + markup — you set a margin from 25% to 200% added to variable usage. Buyers are charged your marked-up amount.