Model pricing and spend¶
ARGUS records what every investigation cost, but it can only do that if you tell it what your model costs. Set prices under Settings → Model pricing.
Why prices are not pre-filled¶
No vendor's rates are built in. A hardcoded price goes stale silently, and a cost figure that is quietly wrong is worse than no figure — so until you configure prices, every analysis records an honest zero.
Setting prices¶
Three prices, in US dollars per million tokens, for the model configured under AI providers:
| Price | Covers |
|---|---|
| Input tokens | Tokens sent to the model |
| Cached-read input tokens | Tokens served from the provider's cache, usually cheaper |
| Output tokens | Tokens the model generated |
All three must be set together. Saving with one or two filled in is rejected, because a partial price table produces a figure that looks complete and is not.
The daily spend ceiling¶
Once prices are set, a check compares today's accumulated spend against the ceiling before starting any automatic investigation. When the ceiling is reached, automatic investigation stops for the rest of the day and says so in the log rather than silently doing nothing. It resumes on its own when spend resets at the next UTC day, or when you raise the ceiling.
A manual Analyze click is a deliberate, bounded human action and is never blocked by the ceiling.
A ceiling needs prices
A ceiling with no price table does nothing, because recorded cost is always zero without prices to compute it from.
Subscription-billed providers¶
An agent-CLI provider authenticated by a subscription rather than per token records cost on a separate, notional basis that the ceiling does not count. With that provider active, real activity cannot breach the ceiling however low it is set. The page says so directly when that is the case.
This is a property of subscription billing rather than something to work around: there is no per-request charge to meter.
Where spend appears¶
The overview shows today's spend against the ceiling. Each analysis records its own cost on the incident. Where a provider cannot report cost, it reads as not measured rather than as zero.