Skip to main content
The Andi AI Search API uses outcome-based pricing. Each call is charged for the work it actually performs, not a flat per-call rate. A simple lookup costs less than a research question that needs several searches and stronger reranking. Calls are charged to your account in the API Console, where you also manage billing.

​
How each endpoint is priced

Endpoint details:

​
What is not charged

  • Failed fetches. 422 and 503 responses from the fetch endpoint are free.
  • Search extracts. extracts=true adds passage-level text to each search result at no additional cost.
  • The news topics index. GET /api/v1/news is free.
News cache hits are charged the per-token rate only, because no search runs.

​
See the cost of each call

Every response reports what it cost, in USD, after any discounts:
  • JSON responses: metrics.cost_dollars
  • format=context responses: cost_dollars in the frontmatter
See metrics for the other fields. For usage totals over time, see the API Console.

​
Control your spend

  • Use auto (the default) for most traffic. Simple queries stay cheap even when your mix includes hard ones.
  • Pin a search mode when you want more uniform cost from call to call. low-cost has a cost ceiling.
  • Cap page text with maxContentLength and limit when you request full content, because returned content is charged per token.
  • Use search extracts=true instead of fetching each result page when passages are enough.
When your balance runs out, requests return 402. See error handling.