Using the AI Chat
The AI Chat panel is the primary interface for Studio AI. This article covers all of its features: the toolbar, composing messages, attaching context, session settings, session history, and more.
Opening AI Chat
Open the AI Chat panel via View > AI Chat or press Ctrl+Alt+I. It appears as a panel on the right side of the Studio.
The AI Chat Toolbar
The toolbar at the top of the AI Chat panel is intentionally minimal. On the chat home screen it shows Browse all chats…, Move View to Secondary Window and More Actions…; when a chat is open, a few more buttons appear.
| Button | Shown | Description |
|---|---|---|
Home (Ctrl+Shift+L) |
In an open chat | Returns to the AI Chat home screen, where a new message starts a new chat. |
Browse all chats... (Ctrl+Alt+L) |
Always | Opens a searchable "Select chat" picker listing your persisted sessions. See Chat Session History below. |
| Summarize Current Session | In an open chat | Creates a summary of the chat so far. |
| Move View to Secondary Window | Always | Pops the AI Chat panel out into its own browser window. |
| Turn Auto Scrolling Off / On | In an open chat | Stops (or resumes) scrolling to the end of the response while it is written. |
| More Actions... (…) | Always | Opens the menu described below. |
The More Actions… menu contains:
| Item | Description |
|---|---|
| Open AI Agent History | Opens the AI Agent History panel in the bottom panel area (alongside Problems/Terminal). See Viewing Token Consumption and AI History. |
| Open AI Configuration | Opens the AI Configuration view as a full editor tab. See AI Configuration: Managing Agents and Settings. |
| Keep screen awake while this chat is working | Shown when a chat is open. Keeps the screen awake while this chat is running. See Long-Running Chats. |
| Set Session Settings... | Opens per-session overrides. See Session Settings below. |
Starting a New Chat
To start a fresh conversation:
- Click Home (
Ctrl+Shift+L) in the chat toolbar to return to the AI Chat home screen, from which typing a new message starts a new session, or - Open the command palette (F1) and run Chat: New Chat.
Session Settings
Set Session Settings… (in the More Actions… menu) opens a dialog with per-session overrides:
- Confirmation Timeout: when enabled, automatically denies tool confirmations after a specified number of seconds, overriding the global preference for this session only.
- Server-Side Compaction: overrides the global and per-provider compaction settings for this session (Compaction mode, Token threshold, and an Advanced Settings JSON block), for providers that support it. See AI Features Settings for how the feature works and where its global and per-provider defaults live.
These settings apply only to the current session and do not persist to new chats.
Composing Messages
Type your question or instruction in the "Ask a question" input at the bottom of the panel. Press Enter to send, or press ↑ to recall a previous message from history.
To address a specific agent, start your message with @AgentName. Once an agent is addressed, it is pinned for the rest of the session, so subsequent messages automatically go to the same agent without the @ prefix. To switch agents, type a new @AgentName. For most tasks you can skip this entirely, see Just Start Asking.
Chat Modes
With the default @k2-assistant, the chat input shows a mode selector with Ask Mode, Plan Mode and Act Mode (default). Ask and Plan are read-only; Act makes changes. The mode applies to the current chat session. See Ask, Plan and Act Modes.
Other agents show their own modes, or none. For example, @Coder offers Edit Mode, Agent Mode and Agent Mode (Next) - see Code Editing with @Coder (Legacy).
Chat Input Controls
Below the input field, the chat input area has:
| Control | Description |
|---|---|
| Attach elements to context | Browse and attach files. This is the paperclip button described in Attaching Context below. |
| @Agent | Opens the agent picker, equivalent to typing @. |
| Toggle Capabilities Configuration | Opens the Generic Capabilities Panel. See Agent Capabilities. |
| Reasoning effort selector | Dropdown (default Auto) with levels Off / Minimal / Low / Medium / High / Auto, controlling how much reasoning effort the model applies to your request. Only the levels the selected model supports are offered. Persists as the session default until changed. |
| Model selector | Dropdown (default Default, mapped to the agent's configured model alias) letting you pick a specific model for the session, for example a Claude Opus, Sonnet, Haiku, or Fable variant, or a model from another connected provider. Once changed, the selection persists as the default for the rest of that chat session (it does not reset per message); switching to a new chat resets it back to Default. The list shows the models selected in the Providers & Models page of AI Configuration. See AI Configuration: Managing Agents and Settings for how aliases map to models. |
Above the input, agents that advertise optional capabilities (such as Shell Execution or GitHub) show toggle chips. See Agent Capabilities.
Attaching Context
The quality of AI responses depends heavily on the context you provide. You can attach relevant files and selections in several ways.
Using Context Variables (Recommended)
Type # in the chat input to see a list of all available context variables. The most commonly used are:
Using context variables inside your prompt lets you also describe why a file is relevant. All files added via variables appear in the context overview below the chat input field.
Drag and Drop / Browse-to-add
You can also drag files from the File Explorer directly into the chat input, or click the paperclip button at the bottom-left of the input area to browse and attach files. You can then type the file name among the project files.
This method does not let you annotate why the file is relevant, but is convenient for quick attachments.
Image Support
You can attach images to your messages. Use the context variable #imageContext or drag an image file into the chat input. Images pasted inline are displayed as thumbnails at their insertion point; images attached via the paperclip button are grouped below the message text. This is useful for sharing diagrams, schema screenshots, or UI mockups with the agent.
Agent Pinning
When you address an agent with @AgentName, it is pinned for the ongoing session. This means you only need to type the agent name once: all subsequent messages in that session go to the same agent automatically. To switch to a different agent, simply type @NewAgentName.
Editing Sent Messages
You can edit a previously sent message in the chat. Click the edit icon next to any previous message, make your changes, and resend.
When you edit and resend a message this way, Studio AI automatically branches the conversation into a new session, preserving the original thread exactly as it was. This makes it easy to explore alternative approaches, for example trying a different prompt phrasing or changing a requirement, without losing your current work. The original session remains accessible in Chat Session History.
Reading Responses
While a response streams in, you can follow the model's thinking live, and a summary of its reasoning stays with the response. On a long response, its header stays at the top of the chat while you scroll.
Finding Text in a Chat
Press Ctrl+F (Cmd+F on Mac) anywhere in the AI Chat panel to open Find in Chat and search the current session. The shortcut stays in the chat: it does not open the editor's find box, even when a file is open.
Starting Chat from the Editor
You can initiate a chat message directly from the editor. Right-click on selected code or a file and choose Ask AI (or similar) from the context menu. This pre-populates the chat input with the selected content as context, saving you the manual step of adding #selectedText.
Chat Session History
Every conversation is automatically saved. Press Ctrl+Alt+L or click the history icon in the toolbar to open Browse all chats…, a searchable "Select chat" picker listing your sessions, each with its title, addressed agent, and last-active time. From here you can:
- Click a session to restore and resume it.
- Rename a session using the pencil icon next to it.
- Delete a session using the × icon next to it.
On the home screen, hovering over a chat's row shows a card with its status (for example Running), its last message, the agent, the number of exchanges and the last activity. The row's icons let you rename the chat, keep the screen awake while it runs (coffee cup), or delete it.
The AI Chat home screen also lists your chats: those currently running under Active at the top, and your most recent sessions in a Restored list, for quick resumption without opening the picker.
After a page reload, the Studio reopens the chat you were in. A chat that was cut off while the assistant was still working is listed as Interrupted and can be resumed with Continue. See Long-Running Chats: Resume and Keep Awake.
For more on AI history and token monitoring, see Viewing Token Consumption and AI History.
Agent Completion Notifications
When an agent finishes a task, particularly when running in the background on a larger delegated task, a notification appears in the Studio. The notification includes the agent name, the session it completed, and a "Show Chat" button that takes you directly to that session to review the result.
You can configure whether each agent sends completion notifications from the AI Configuration view.
You can also control overall behavior via Settings > Ai-features > Notifications: Default (Default is Off)
Using the AI Chat
The AI Chat panel is the primary interface for Studio AI. This article covers all of its features: the toolbar, composing messages, attaching context, session settings, session history, and more.
Opening AI Chat
Open the AI Chat panel via View > AI Chat or press Ctrl+Alt+I. It appears as a panel on the right side of the Studio.
The AI Chat Toolbar
The toolbar at the top of the AI Chat panel is intentionally minimal. On the chat home screen it shows Browse all chats…, Move View to Secondary Window and More Actions…; when a chat is open, a few more buttons appear.
| Button | Shown | Description |
|---|---|---|
Home (Ctrl+Shift+L) |
In an open chat | Returns to the AI Chat home screen, where a new message starts a new chat. |
Browse all chats... (Ctrl+Alt+L) |
Always | Opens a searchable "Select chat" picker listing your persisted sessions. See Chat Session History below. |
| Summarize Current Session | In an open chat | Creates a summary of the chat so far. |
| Move View to Secondary Window | Always | Pops the AI Chat panel out into its own browser window. |
| Turn Auto Scrolling Off / On | In an open chat | Stops (or resumes) scrolling to the end of the response while it is written. |
| More Actions... (…) | Always | Opens the menu described below. |
The More Actions… menu contains:
| Item | Description |
|---|---|
| Open AI Agent History | Opens the AI Agent History panel in the bottom panel area (alongside Problems/Terminal). See Viewing Token Consumption and AI History. |
| Open AI Configuration | Opens the AI Configuration view as a full editor tab. See AI Configuration: Managing Agents and Settings. |
| Keep screen awake while this chat is working | Shown when a chat is open. Keeps the screen awake while this chat is running. See Long-Running Chats. |
| Set Session Settings... | Opens per-session overrides. See Session Settings below. |
Starting a New Chat
To start a fresh conversation:
- Click Home (
Ctrl+Shift+L) in the chat toolbar to return to the AI Chat home screen, from which typing a new message starts a new session, or - Open the command palette (F1) and run Chat: New Chat.
Session Settings
Set Session Settings… (in the More Actions… menu) opens a dialog with per-session overrides:
- Confirmation Timeout: when enabled, automatically denies tool confirmations after a specified number of seconds, overriding the global preference for this session only.
- Server-Side Compaction: overrides the global and per-provider compaction settings for this session (Compaction mode, Token threshold, and an Advanced Settings JSON block), for providers that support it. See AI Features Settings for how the feature works and where its global and per-provider defaults live.
These settings apply only to the current session and do not persist to new chats.
Composing Messages
Type your question or instruction in the "Ask a question" input at the bottom of the panel. Press Enter to send, or press ↑ to recall a previous message from history.
To address a specific agent, start your message with @AgentName. Once an agent is addressed, it is pinned for the rest of the session, so subsequent messages automatically go to the same agent without the @ prefix. To switch agents, type a new @AgentName. For most tasks you can skip this entirely, see Just Start Asking.
Chat Modes
With the default @k2-assistant, the chat input shows a mode selector with Ask Mode, Plan Mode and Act Mode (default). Ask and Plan are read-only; Act makes changes. The mode applies to the current chat session. See Ask, Plan and Act Modes.
Other agents show their own modes, or none. For example, @Coder offers Edit Mode, Agent Mode and Agent Mode (Next) - see Code Editing with @Coder (Legacy).
Chat Input Controls
Below the input field, the chat input area has:
| Control | Description |
|---|---|
| Attach elements to context | Browse and attach files. This is the paperclip button described in Attaching Context below. |
| @Agent | Opens the agent picker, equivalent to typing @. |
| Toggle Capabilities Configuration | Opens the Generic Capabilities Panel. See Agent Capabilities. |
| Reasoning effort selector | Dropdown (default Auto) with levels Off / Minimal / Low / Medium / High / Auto, controlling how much reasoning effort the model applies to your request. Only the levels the selected model supports are offered. Persists as the session default until changed. |
| Model selector | Dropdown (default Default, mapped to the agent's configured model alias) letting you pick a specific model for the session, for example a Claude Opus, Sonnet, Haiku, or Fable variant, or a model from another connected provider. Once changed, the selection persists as the default for the rest of that chat session (it does not reset per message); switching to a new chat resets it back to Default. The list shows the models selected in the Providers & Models page of AI Configuration. See AI Configuration: Managing Agents and Settings for how aliases map to models. |
Above the input, agents that advertise optional capabilities (such as Shell Execution or GitHub) show toggle chips. See Agent Capabilities.
Attaching Context
The quality of AI responses depends heavily on the context you provide. You can attach relevant files and selections in several ways.
Using Context Variables (Recommended)
Type # in the chat input to see a list of all available context variables. The most commonly used are:
Using context variables inside your prompt lets you also describe why a file is relevant. All files added via variables appear in the context overview below the chat input field.
Drag and Drop / Browse-to-add
You can also drag files from the File Explorer directly into the chat input, or click the paperclip button at the bottom-left of the input area to browse and attach files. You can then type the file name among the project files.
This method does not let you annotate why the file is relevant, but is convenient for quick attachments.
Image Support
You can attach images to your messages. Use the context variable #imageContext or drag an image file into the chat input. Images pasted inline are displayed as thumbnails at their insertion point; images attached via the paperclip button are grouped below the message text. This is useful for sharing diagrams, schema screenshots, or UI mockups with the agent.
Agent Pinning
When you address an agent with @AgentName, it is pinned for the ongoing session. This means you only need to type the agent name once: all subsequent messages in that session go to the same agent automatically. To switch to a different agent, simply type @NewAgentName.
Editing Sent Messages
You can edit a previously sent message in the chat. Click the edit icon next to any previous message, make your changes, and resend.
When you edit and resend a message this way, Studio AI automatically branches the conversation into a new session, preserving the original thread exactly as it was. This makes it easy to explore alternative approaches, for example trying a different prompt phrasing or changing a requirement, without losing your current work. The original session remains accessible in Chat Session History.
Reading Responses
While a response streams in, you can follow the model's thinking live, and a summary of its reasoning stays with the response. On a long response, its header stays at the top of the chat while you scroll.
Finding Text in a Chat
Press Ctrl+F (Cmd+F on Mac) anywhere in the AI Chat panel to open Find in Chat and search the current session. The shortcut stays in the chat: it does not open the editor's find box, even when a file is open.
Starting Chat from the Editor
You can initiate a chat message directly from the editor. Right-click on selected code or a file and choose Ask AI (or similar) from the context menu. This pre-populates the chat input with the selected content as context, saving you the manual step of adding #selectedText.
Chat Session History
Every conversation is automatically saved. Press Ctrl+Alt+L or click the history icon in the toolbar to open Browse all chats…, a searchable "Select chat" picker listing your sessions, each with its title, addressed agent, and last-active time. From here you can:
- Click a session to restore and resume it.
- Rename a session using the pencil icon next to it.
- Delete a session using the × icon next to it.
On the home screen, hovering over a chat's row shows a card with its status (for example Running), its last message, the agent, the number of exchanges and the last activity. The row's icons let you rename the chat, keep the screen awake while it runs (coffee cup), or delete it.
The AI Chat home screen also lists your chats: those currently running under Active at the top, and your most recent sessions in a Restored list, for quick resumption without opening the picker.
After a page reload, the Studio reopens the chat you were in. A chat that was cut off while the assistant was still working is listed as Interrupted and can be resumed with Continue. See Long-Running Chats: Resume and Keep Awake.
For more on AI history and token monitoring, see Viewing Token Consumption and AI History.
Agent Completion Notifications
When an agent finishes a task, particularly when running in the background on a larger delegated task, a notification appears in the Studio. The notification includes the agent name, the session it completed, and a "Show Chat" button that takes you directly to that session to review the result.
You can configure whether each agent sends completion notifications from the AI Configuration view.
You can also control overall behavior via Settings > Ai-features > Notifications: Default (Default is Off)
