AI Assistant

A chat assistant in the side panel. Ask a question, paste or attach an image, and it answers — and when you want, it can act on your project: generating artifacts and running multi-step tasks for you.

What you can do

• Chat with an LLM from the side panel, including questions about images you paste or attach.
• Generate a project artifact from a prompt: the assistant drafts the file, fills it in, checks it, and opens it for you. Finished artifacts show as a card with an "Open" button.
• Let it plan and run a multi-step task, watching each step's status as it goes.
• Review what it wants to do before it does it: pending actions appear inline to approve once, for this workspace, always, or dismiss.
• Speak instead of typing: press the microphone beside Send, and what you say lands in the message box on your own machine (needs the Whisper Runtime extension).
• Hear the answer: a message you dictated is answered aloud by the Kokoro Voice extension, or by your system's voice where it cannot run.

Settings

Use the settings popover in the panel to choose the active model and to switch between suggesting actions and running them automatically.

The language you dictate in and whether answers are read aloud are the Whisper Runtime extension's Voice settings, shared with the Agent window.

Getting started

Open the AI Assistant from its icon in the secondary sidebar, then type a prompt.