> ## Documentation Index
> Fetch the complete documentation index at: https://docs.mossx.ai/llms.txt
> Use this file to discover all available pages before exploring further.
# Chat area
> Read streamed replies, thinking, tool calls, context usage, and the status panel.
The chat area is the main CC GUI surface. Here you see how the assistant thinks, which tools it called, which files it changed, and how much context this turn used.
## Empty sessions and the welcome screen
When a new session has no messages, the middle shows a welcome screen: the current provider's logo, version, and "Send a message to …". Click the version to open the changelog.
After the SDK is installed and a provider is enabled, the input box is ready. If it says the SDK is not installed, click **Go to install** to jump to [SDK dependencies](/en/vscode/start/setup).
## What a message can contain
Assistant messages are usually several blocks:
| Block | Meaning |
| ---------- | -------------------------------------------------------------------------------------------------------- |
| Thinking | The model's internal reasoning. You can collapse it |
| Body | Markdown reply. Code blocks can be copied |
| Tool calls | Read files, edit files, run commands, search, open pages |
| Diff | File change preview. Click to open the file |
| Usage | Input / output tokens for this turn. With detailed output on, you also see cache hits and estimated cost |
A spinner shows while a tool call is running. It should stop after a multi-step task ends. If it keeps spinning, see [FAQ](/en/vscode/community/faq).
Click a message to roll back to before that turn. See [Sessions, tabs, and history](/en/vscode/chat/sessions#roll-back-to-a-message).
## Streaming and stop
Streaming is on by default: text and tool blocks appear as they are generated. Turn it off in **Settings → General → Behavior** to receive each reply as a whole.
While generating, the input button becomes **Stop**. Click it to interrupt the current turn:
* Claude / Codex: end the current request
* Grok / Kimi / OpenCode / Pi: also end the matching child process
If you stop it yourself, the completion toast and sound do not play.
## Search in the chat
Press `Cmd+F` (Windows / Linux: `Ctrl+F`) to search inside the conversation.
* `Enter` / `Shift+Enter`: next / previous
* `Alt+C`: match case
* `Alt+W`: whole word
* `Alt+R`: regular expression
* `Esc`: close
If early messages are collapsed, search expands them first, then jumps to the match.
## Context usage
The context indicator near the input box shows how much of the current window is used. Click it for a finer breakdown.
When you are close to the limit you can:
* Send `/compact` so the model summarizes earlier turns
* Start a new session and keep only the files you need
* Avoid pasting large logs. Mention files with `@` instead
After `/compact`, you can expand or collapse the summary node. Messages before the summary stay in history. They are just collapsed by default.
## Status panel
On longer tasks, a status panel appears on the right or at the bottom:
| Tab | Contents |
| ------------- | ------------------------------------------------------------- |
| Todo | Tasks the model split out |
| Subagents | Parallel subtasks |
| Files / edits | Files changed this turn. You can view the diff, undo, or keep |
Undoing a file change cannot be undone again. Review the list before you confirm.
## Permissions, questions, and Plan
When the model needs to edit a file, run a command, or ask you a multiple-choice question, a dialog appears. Set the timeout in **Settings → General → Behavior**, from 30–3600 seconds. After timeout the dialog closes and the model continues with the current policy.
In plan mode, the model first gives a read-only plan and waits for your approval before it edits code. See [Modes and permissions](/en/vscode/chat/modes).
## Related pages
`@`, `/`, `#`, `!`, attachments, and send shortcuts.
Tabs, stars, export, and rollback.
