A credit is roughly a few cents of LLM spend in Convira's cloud. A short cloud Q&A burns a fraction of a credit. A long agentic run that calls many tools and reasons over a big context can burn dozens.
The cost depends on the model and the prompt - Convira shows you the per-run cost in the dashboard so you can see what you're spending in real time.
Local AI and Private Box inference don't consume credits. When the model runs on your hardware, there is no model usage for us to meter. Optional online services such as web research, media generation, and connector actions can still use credits when you invoke them on an eligible plan.