Kilo Code v7.3.46 Changelog

This release is all about giving you finer control over AI helpers and making your daily workflow smoother. It matters most to developers who use custom AI setups and want to prevent weird glitches when managing their code.

What's new

• Custom AI thinking for helper bots: You can now tell specific "subagents" (smaller AI helpers that do focused tasks) exactly which AI model to use and how much "reasoning" (step-by-step thinking) they should do.
• Turn off reasoning later: If you turn on the step-by-step reasoning feature for a custom AI model, you can now turn it back off if you change your mind.

Improvements

• Clearer permission pop-ups: When the AI asks for permission to run a repeating task (a "loop"), the pop-up warnings are now clearer and use tool names that match your language, making them easier to understand.
• Smarter question routing: When you answer a question the AI asks you, your answer now goes straight to the exact "worktree" (a separate, temporary copy of your project folder) where the question was originally asked. This keeps things organized.

Fixes

• Fixed the Redo button: If you used "Undo" to go back to a previous chat session, clicking "Redo" wasn't always bringing you back to where you just were. This is now fixed so it restores your session on the very first click.
• Stopped accidental folder deletion: Previously, if you removed a "skill" (a specific ability or tool) from the AI, it would accidentally trigger a chain reaction that deleted your entire working directory folder. That dangerous bug is squashed.

Breaking changes

No breaking changes in this release.