Auto mode, voice input, and file attachments
v3.6.0 adds a new Auto mode that reads your project and picks the next step on its own, while you still approve anything that sends, deletes, or touches a secret. Modes are now Ask, Plan, Build, and Auto, and you can dictate prompts by voice, attach files to chat, and review every change the agent makes in one place.
- New Auto mode reads your project, picks the next step, and verifies its work, while you still approve anything that sends, deletes, or touches a secret (Plus, Pro, and Lifetime plans)
- Revert an entire Auto turn in one click if you don't like the result
- Work modes renamed to Ask, Plan, Build, and Auto, with a new mode picker in the prompt bar
- Voice dictation: record a prompt and have it transcribed using a supported transcription provider and model from Settings
- Attach files (not just images) to chat messages, plus drag-and-drop text into the prompt bar
- New file changes summary shows every file the agent edited or created, including pending diffs awaiting review
- Smarter project context: the agent analyzes your project structure to pull in the right files, tools, and skills for each request
- Chat messages show which skills were applied, with clearer grouped tool activity and findings
- Faster multi-file reads, localized voice error messages, a corrected light/dark theme toggle icon, and assorted stability fixes