An AI API spending limit is a dollar cap you set with your AI provider. ExactFind can't set one for you, because it only counts tokens. This limit is the one that protects your bill.
Alerts and hard limits
| Type | What it does |
|---|---|
| Alert | Tells you when spending reaches an amount, and requests carry on |
| Hard limit | Stops requests when spending reaches an amount |
Check which one you've set. They can sit next to each other in a provider's settings.
Claude (Anthropic)
Set your limit on the Billing page of the Claude Console. You can also set one for each workspace.
ChatGPT (OpenAI)
Go to your organization's limits and find Spend.
Turn on Enforce a hard limit. Without it, the limit only sends alerts and doesn't stop traffic. Enforcement isn't instant, so spending can go slightly over.
OpenAI-compatible services
Each service has its own billing, so check whether yours offers a hard limit.
Use a separate key for each site
Create an API key just for this site. If you use a workspace or project for it, you can see its spending on its own and set a limit that doesn't affect anything else.
If you use different platforms for answers and indexing, set a limit on both.
How this fits with the token limits
| Limit | Where | What it counts |
|---|---|---|
| Spending limit | Your provider's account | Dollars |
| Monthly token limits | ExactFind | Tokens, in four kinds |
Use both. The token limits let ExactFind show visitors a fallback link. The spending limit covers everything else on the account.
When a hard limit is reached
Your provider starts refusing requests, and visitors see "Answers aren't available right now". See Fix “Answers Aren't Available Right Now” Errors.
Provider settings change, so check their current documentation when you set this up.