Your model provider bill (OpenAI, Anthropic, Google, and others) is separate and paid to the provider. Control that spend with Budgets and per-key limits.
Plans
There is no per-seat charge. Usage resets on the first of each month, UTC.
Metered products
Pricing is graduated: each unit is priced at the rate of the tier it falls in, so only usage above a threshold gets that tier’s price.
LLM observability (per event)
LLM observability (per event)
Agent runs (per minute)
Agent runs (per minute)
Routing (per decision)
Routing (per decision)
Read your usage
On Pay-as-you-go, the top of the page shows:- This month: charges so far in the current period.
- Projected: this month’s usage carried to month end at the same pace, priced on each product’s tiers.
Set billing limits
A billing limit is a dollar ceiling per product for the month.1
Open Billing
Find the product card you want to cap.
2
Edit the billing limit
Enter a dollar amount, or choose no limit. Agent runs start with a default limit of $150.
3
Save
Save the limit. The card shows the new ceiling immediately.
Packages
Packages add governance and identity features on top of your plan.
Add or remove Boost and Scale on the Billing page. If you remove a package, controls you already configured stay in place, but you cannot create new ones that need the package. For Enterprise, contact Caveman from the Billing page.
Your included AI
Every user gets a monthly AI credit allowance (currently $10) for Caveman’s built-in agents and automatic model decisions. It is personal and shared across all organizations you belong to, valued at catalog list prices, and renews monthly.- AI credits remaining and AI credits used show your balance. In-flight requests reserve credits until their cost is known.
- Caveman uses Luna for routine work and Sol for complex tasks.
- When credits run out, Caveman’s AI work pauses. Your own application traffic is unaffected and keeps using your provider billing.