Credits & Usage
QueryBurst Answer Engine uses a credit-based model. One processed conversation turn uses one credit. Each turn involves multiple model operations, web searches, retrieval, scoring, and grounding — substantially more work than a single API call.
What is an Answer Engine credit?
One processed conversation turn uses one Answer Engine credit.
So a 10-turn conversation uses 10 credits.
Creating personas, journeys, and branches does not itself consume processing credits. Credits are used when messages are actually run through the research pipeline.
Why does each message have a meaningful cost?
A turn is not a single model call.
It can involve:
- multiple model operations
- fan-out generation
- multiple web searches
- retrieval and extraction
- passage scoring
- entity and relationship analysis
- grounding
- final generation
And as a conversation grows, later turns also need the relevant state from earlier turns.
Answer Engine is therefore doing substantially more work than firing one prompt at an API endpoint.
Do I pay again when I branch a conversation?
Only when you process new messages on that branch.
Creating the branch itself is simply creating another possible path.
Each processed turn uses a credit.
Can I queue journeys and let them run later?
Longer runs can be queued rather than requiring you to sit and watch every message complete interactively.
This is particularly useful when running multiple personas or larger sets of journeys.
Do unused Answer Engine credits roll over?
Currently, Answer Engine allowances reset with each billing period.
Unused credits do not currently accumulate.
That may change as usage patterns become clearer.
Can I export Answer Engine data?
Not yet.
Export functionality is planned, but the current release focuses on interactive inspection inside Answer Engine.