Chat Interface
Everything you need to know about the conversation interface.
Chat and Work Modes
The conversation interface operates in two modes. Chat mode is for straightforward question-and-answer exchanges. Work mode is for multi-step tasks where the agent plans, executes, and reports progress autonomously.
| Chat | Work | |
|---|---|---|
| Best for | Quick questions, translations, summaries | Code generation, research, multi-file edits |
| Agent behavior | Responds to each message individually | Breaks the task into steps, executes them in sequence |
| Progress display | None | Progress panel with step-by-step status |
Click the mode button above the input bar → select the desired mode.
[!NOTE] Screenshot placeholder.
Input Bar
The input bar is where you compose messages and control how the agent processes them.
Persona and Model Selection
Different personas give the agent different behavior and expertise. Different models trade speed for capability. Click the PersonaModelBar above the input bar → choose a persona → choose a model.
[!NOTE] Screenshot placeholder.
Slash Commands
Typing / in the input bar triggers slash commands:
| Command | Action |
|---|---|
/plan | Enter plan mode — the agent outlines its approach before executing |
/setup | Open project settings |
File Attachments
Attach files for the agent to read. Drag a file onto the input bar, or click the attachment button to browse.
Voice Input
Click the microphone button to dictate your message. This requires ASR credentials to be configured in Settings.
Reasoning Effort
Reasoning effort controls how deeply the agent thinks before responding. Higher effort produces more thorough answers but takes longer.
| Level | Behavior |
|---|---|
| Off | No reasoning step, fastest response |
| Minimal | Brief reasoning, good for simple tasks |
| Low | Short reasoning chain |
| Medium | Balanced depth and speed |
| High | Extended reasoning for complex problems |
| Max | Deepest reasoning, longest wait |
Set the reasoning effort via the control in the input bar.
Auto-resizing Input
The text input expands as you type and collapses back when empty. For long messages, it grows up to a maximum height and then scrolls internally.
Message Actions
Each message in a conversation offers a set of actions.
Copy and Edit
Copy a message to your clipboard via the copy button on the message. Edit a previously sent message by clicking the edit button → modify the text → resend. The agent will regenerate its response from that point.
Report a Message
If a response is problematic, click the report button → choose a reason:
| Reason | Meaning |
|---|---|
| Inaccurate | Contains factual errors |
| Out of date | Information is no longer correct |
| Too short | Response is insufficiently detailed |
| Too long | Response is unnecessarily verbose |
| Harmful | Content is dangerous or offensive |
| Wrong sources | Citations do not support the claims |
Star and Pin
Star a conversation to mark it as important. Pin a conversation to keep it at the top of the sidebar. Both are toggles — click again to undo.
Incognito Mode
Incognito conversations are not saved to history and are permanently deleted after 24 hours. Enable incognito mode when you want a private exchange without leaving a trace.
Share, Rename, Delete
Share a conversation to generate a link others can view. Rename a conversation to make it easier to find later. Delete a conversation to remove it permanently. All three actions are available from the conversation header menu.
Interrupt and Retry
If the agent is taking too long or heading in the wrong direction, click the stop button to interrupt. Click retry to regenerate the last response with a fresh sample.
Cop Timeline
The cop timeline is the execution trace shown inside each AI reply. It reveals what the agent did behind the scenes to produce its answer.
Thinking Blocks
When reasoning effort is enabled, the timeline shows a thinking block with a duration counter and a rotating hint such as "Thinking..." or "Analyzing code...". Click to expand and read the reasoning.
Tool Calls
Each tool invocation appears as a card in the timeline:
| Tool | Display |
|---|---|
| Shell execution | Command, output, and success/failure status |
| Web fetch | Retrieved content summary |
| Code execution | Output from the sandbox |
Sources
When the agent cites external material, a source list appears at the bottom of the reply. Each source links to the original content.
Expand and Collapse
Long timelines start collapsed. Click "Show more" to expand the full trace, or "Show fewer" to collapse it again.
[!NOTE] Screenshot placeholder.
Special Cards
Certain actions and results render as rich content cards inside the conversation.
| Card | Purpose |
|---|---|
| TodoListCard | Track task progress with checkable items |
| CodeExecutionCard / CodeExecutionPanel | Display sandbox code execution output |
| SubAgentBlock | Show sub-agent status: Spawning, Running, Completed, Failed, Closed |
| Artifact preview | Render HTML, SVG, image, or stream blocks inline |
| UserInputCard | Agent requests additional input from you mid-task |
| MemoryActionBlock | Show memory operations: write, search, read, delete |
| MermaidBlock | Render Mermaid diagrams |
| MindmapBlock | Render mind maps |
| GeoGebraBlock | Render GeoGebra interactive math figures |
| ContextCompactBar | Show automatic context compression progress |
[!NOTE] Screenshot placeholder.
Right Panel
The right panel is a switchable area next to the conversation. Toggle between panels using the tabs at the top.
Browser Panel
A built-in browser for previewing web pages. Navigate with forward, back, and refresh controls. Manage bookmarks and view browsing history.
Files Panel
A local file browser. Navigate your project directory and open files for the agent to reference.
Resource Panel
Preview resources such as Markdown documents, source code, and original reference files.
[!NOTE] Screenshot placeholder.
Sidebar
The sidebar lists all your conversations and provides navigation.
Normal Mode
Conversations are listed in chronological order, most recent first.
GTD Mode
Conversations are organized into action groups:
| Group | Meaning |
|---|---|
| Backlog | Not yet started |
| Next | Ready to work on |
| Waiting | Blocked on something external |
| Someday | Deferred indefinitely |
| Archived | Completed and stored |
Drag a conversation between groups to reclassify it.
Pinned Section
Pinned conversations appear in a separate section at the top of the sidebar for quick access.
Project Groups
When a conversation belongs to a project, it is nested under that project's heading. Collapse or expand a project group by clicking its header.
Search and New Conversation
Use the search field at the top of the sidebar to find a conversation by name. Click the new conversation button to start a fresh thread.
[!NOTE] Screenshot placeholder.