docs/workspaces/chat-interface.mdx
The chat interface is your primary way of communicating with coding agents. It supports rich message formatting, file attachments, approval workflows, and multiple conversation sessions.
The chat displays different types of messages, each with distinct styling:
Your messages appear with an edit option. Click the pencil icon to modify and resend a message.
Agent responses are rendered with full markdown support, including:
Informational messages from the system, such as model initialisation or status updates.
Error messages appear in red with a warning icon. Click to expand and see full error details.
The chat input supports rich text editing:
Cmd/Ctrl + B or click BCmd/Ctrl + I or click ICmd/Ctrl + U or click UYou can attach images to your messages:
Type @ to mention files from your repository. The chat provides typeahead suggestions showing:
Select a file to include it as context for the agent.
| Action | Shortcut | Description |
|---|---|---|
| Send | Cmd/Ctrl + Enter | Send message immediately |
| Queue | Click Queue button | Queue message while agent is running |
| Stop | Click Stop button | Stop the current agent execution |
The chat interface shows different states:
| State | Description |
|---|---|
| Idle | Ready to send a new message |
| Running | Agent is actively working |
| Queued | Your message is queued for when agent finishes |
| Sending | Message is being sent |
Select which coding agent to use from the Agent dropdown in the chat toolbar. Available agents depend on your configuration.
Some agents support different variants or profiles. Use the variant selector to choose:
Click Customise to configure agent settings.
Some agents (especially with planning mode enabled) ask for your approval before making changes. This gives you control over what gets implemented.
When the agent needs approval, you'll see an approval card in the chat:
If the plan isn't quite right:
If you want the agent to work autonomously without asking for approval:
dangerously_skip_permissions in agent settings (use with caution)You can edit and resend previous messages:
<Frame> </Frame> <Steps> <Step title="Click the Edit Button"> Click the pencil icon on any of your messages. </Step> <Step title="Modify Your Message"> The message content appears in an editable text area. Make your changes. <Frame> </Frame> </Step> <Step title="Submit the Edit"> Click **Retry** to resend the modified message. The conversation continues from that point. Click **Cancel** to discard your changes. </Step> </Steps> <Warning> Editing a message creates a new branch in the conversation. Subsequent messages after the edited one will be replaced. </Warning>The context gauge shows how much of the agent's context window is used. Understanding this helps you get better results.
Tokens are how AI models measure text. Roughly:
The context window includes everything the agent "remembers": your messages, its responses, code it has read, and file contents.
When usage is high:
| Colour | Usage | What to Do |
|---|---|---|
| Grey | 0-30% | All good, continue working |
| Default | 30-60% | Normal usage, keep going |
| Orange | 60-80% | Consider starting a new session soon |
| Red | 80%+ | Start a new session - agent is near its limit |
When the agent breaks down work into tasks, a progress indicator appears in the chat toolbar:
Click the indicator to expand the full task list showing each step the agent is working through.
The chat shows file modification summaries:
When the agent runs commands, you'll see:
Click to view full output in the logs panel.
Search operations show summarised results. Click to expand and see full search output.
When you add inline comments in the Changes panel:
| Shortcut | Action |
|---|---|
Cmd/Ctrl + Enter | Send message (or contextual action) |
Cmd/Ctrl + B | Bold text |
Cmd/Ctrl + I | Italic text |
Escape | Cancel current action |