Investigation Cost More Than Expected

A few things can cause a run to consume more credits than typical:

A higher model level or effort

Credits are charged on what the turn consumed, priced at the model that served it, so a more capable level or a higher reasoning effort costs more per turn. Each answer reports the level and effort it ran with. Check the picker in the Chat prompt toolbar to see where the thread is now, and Settings → Agent0 to see the default it inherited. See Model Selection.

Multi-step workflows

A request that triggers RCA, code analysis, and PR generation consumes credits for all three steps. This is intended. Use the tool call transparency in the conversation to see what each step cost.

Repeated tool calls during investigation

When Agent0 forms and tests hypotheses, it may run multiple queries against your telemetry. Each query consumes tokens. This is how investigations work. The alternative is shallow guesses that don't reach root cause.

Large repository clones

If Agent0 needs to clone a large repository to investigate a code-level issue, the analysis consumes more tokens than a query-only investigation. The first investigation against a given repository is typically the most expensive; subsequent runs benefit from cached context.

Further Reading

• About Troubleshooting — Overview of troubleshooting topics
• Investigation & Analysis — Understanding how investigations work
• Key Concepts — Understand credits and runs