Kilo Code v7.5.0 Changelog

This release is all about making the Agent Manager smarter and less cluttered. It matters most to developers who use Kilo Code to run multiple AI coding assistants at once, giving them better ways to keep track of background tasks, review AI-suggested code changes, and manage their AI usage limits without leaving their workflow.

What's new

• Background AI helpers are now on by default: You can let AI "subagents" (smaller AI tasks) run in the background while you keep working. They show up in a collapsible bar at the top of your chat so you don't lose track of them.
• Answer AI questions directly: If an AI helper is stuck waiting for an answer, the main "orchestrating" AI can now answer that question for it instead of just stopping the whole task.
• Check your AI usage: You can now see how much of your AI plan you've used and your quota limits right in the command line or VS Code settings.
• Document Inspector: You can now preview Markdown files (like read-me docs) with review comments right inside the Agent Manager, while regular code files open in your normal editor.
• Copy PR links: There's a new button to easily copy web links for Pull Requests (proposals to merge new code) directly from the Agent Manager panel.
• Better project history: The sidebar no longer shows a massive list of sessions. Instead, there's a single history button for each project, making it much easier to find and resume past work.

Improvements

• Cleaner file changes: When the AI edits files, those changes now open right in the Agent Manager side panel instead of cluttering your screen with new tabs. You can click to expand them, and multi-file changes open all at once with the right sizing.
• Faster worktrees: Creating a "worktree" (a separate copy of your code to test changes in) is now much faster, and ready sessions pop up immediately. Loading code differences (diffs) for these worktrees is also speedier.
• Tidier UI: The Agent Manager toolbar is now evened out so all buttons share the same height and spacing. Also, the chat input box no longer jumps around when a task finishes.
• Better chat history: Answered questions in your chat history now stay expanded and wrap their text properly so it's easier to read.
• Smarter model searches: When searching for a specific AI model, you can now search by the provider's name (like "openai:gpt-4") and it will actually find it.
• Terminal stability: The built-in terminal now uses a more reliable way to draw text on your screen, preventing graphics glitches, and it stops wasting computer power rendering terminals you can't even see.
• Smoother multi-tasking: When running multiple AI sessions at once, the app now processes events more efficiently without getting bogged down. Foreground subagents can also continue in the background independently.

Fixes

• Stuck sessions: Fixed a bug where finished AI sessions would get stuck showing a "working" spinner forever.
• Duplicate text: Fixed an issue where the AI's reasoning and response text would accidentally show up twice while subagent sessions were open.
• Project switching bugs: Fixed several bugs when switching Agent Manager projects, including false errors saying GitHub wasn't installed, and leftover "stale" states messing up the history.
• Glitchy event tracking: Fixed a bug where the system trying to prevent duplicate events accidentally hid real, delayed updates after reconnecting or during busy bursts of activity.
• Keyboard shortcuts: Fixed the VS Code command picker so that pressing Enter or Tab actually runs the slash command you highlighted.
• Onboarding settings: Fixed an issue so that your permission settings for auto-approving code remain editable during the review-first onboarding step.

Breaking changes

No breaking changes in this release.