Your First Chat

The chat panel has three tabs at the top:

• CHAT -- The main conversation view. This is where you type messages and see AI responses.
• OUTPUT -- A scrollable log of all conversation history, including raw API requests and responses. Useful for debugging.
• TIMELINE -- Shows checkpoints created during your session. You can restore, diff, or branch from any checkpoint.

Below the tabs is the message area where the conversation appears, and at the bottom is the input field where you type your messages.

The chat header displays several important indicators:

• Mode Selector -- Shows the current mode (Agent, Ask, Plan, Edit, Apps Builder, or PRD). Click to switch modes.
• Model Indicator -- Displays the currently selected AI model. Click to change models.
• Cost Indicator -- Shown in lime green, this tracks the running cost of API calls for the current session. Resets when you clear the conversation.
• Context Bar -- A thin colored bar that shows how much of the model's context window you have used:
• Green -- Plenty of context remaining (under 50%).
• Yellow -- Getting full (50-80%). Consider clearing old messages.
• Red -- Near the limit (over 80%). The AI may start losing context from earlier messages.
• Sandbox Indicator -- Shows whether Docker sandbox mode is active. When enabled, all file operations and commands run inside a container.
• Agent Indicator -- Shows which sub-agent is currently active (Coder, Reviewer, Planner, etc.) when multi-agent routing is enabled.

Type your message in the input field at the bottom of the chat panel and press Enter to send.

You can ask QCoder to: - Explain code: *"What does this function do?"* - Write code: *"Write a function that sorts an array of objects by date"* - Fix bugs: *"Why is this test failing?"* (with the relevant file open) - Refactor: *"Refactor this class to use the strategy pattern"* - Execute commands: *"Run the test suite and fix any failures"* (in Agent mode)

The AI response appears in the message area with full markdown rendering, syntax-highlighted code blocks, and interactive elements.

When the AI uses a tool (reading a file, writing code, running a command), it appears as a tool call card in the conversation:

Each card shows: - Tool name -- e.g., read_file, write_to_file, execute_command - Status -- A spinner while running, a checkmark when complete, or an error icon if it failed. - Parameters -- The arguments passed to the tool (e.g., the file path, the command to run). - Result preview -- A collapsible section showing the tool's output (e.g., file contents, command output).

In Agent mode, tools execute automatically. In Ask mode, the AI will describe what it would do but will not execute tools. In other modes, you may be prompted to approve certain operations.

When the AI responds with code, it appears in syntax-highlighted code blocks. If the code is a suggested edit to an existing file, you will see action buttons:

• Apply -- Applies the suggested changes to the file, showing a diff overlay with green (added) and red (removed) lines.
• Copy -- Copies the code block to your clipboard.
• Insert at Cursor -- Inserts the code at your current cursor position in the editor.

After applying changes, you can use Accept to confirm or Reject to undo.