What it does
Audit mode records suggestions without changing your settings. Failed or uncertain scores leave effort alone; the savings display uses assumed ratios, not measured token savings.
Agent tooling
Jeffort
Choose Claude Code's effort level from the prompt before the larger model starts work.
Bookmark: Jeffort
Keep this in your collection.
Leave a noteWhat would you try with this?
: Jeffort
Only you can see your notes.
Screenshot unavailable. Open the experiment ↗
How you can use it
If you use the Claude Code coding assistant, start by watching Jeffort's suggestions. Its audit mode records how much effort it would choose, but leaves your setting alone. Compare those suggestions on small and difficult jobs before letting it change the setting.
Your requests go to TypeSafe's outside AI service for scoring. You need paid credits and a service access key. The savings display is an estimate; it does not prove that using less effort preserves the quality of difficult answers.