Skip to main content
The message composer is the bar at the bottom of every chat. It is where you type a question, pick which AI agent answers, choose how hard it should think, and decide which tools and data it is allowed to use for the next message you send.

The composer at a glance

Everything you set here applies to the chat you are currently in, and it is remembered per chat. Anything you have turned on shows up as a small removable pill (chip) just above the text box.

What you can do here

Control / ButtonWhat it does
Message box (“Ask about your data…”)Type your question or instruction here.
EnterSends the message.
Shift + EnterAdds a new line without sending.
Send button (arrow icon)Sends the message.
Stop button (square icon)Appears only while a reply is being generated; cancels the in-progress response.
Deep Thinking button (brain icon)A three-step toggle: “Deep Thinking”, “2X Deep Thinking”, or off.
Attach file button (paperclip icon)Opens the chat’s Attachments (artifacts) panel.
Web Search toggleTurns on web search for this message.
Agent picker (robot icon)Choose which AI agent answers. Lists the agents available to your department.
Tools button (wrench icon, with an “X/Y” count)Opens a list of the optional tools the chosen agent can use for this message.
Reset (inside the Tools popover)Clears any tool changes and returns to the agent’s normal tool set.
Active chips (above the text box)Pills showing what’s switched on. Click the X on a pill to remove that setting.

What “agent mode” means

“Mode” is simply how the agent works through your request, and the Deep Thinking button is how you choose it:
  • Off (default): A fast, conversational mode that can still use tools to answer most everyday questions.
  • Deep Thinking: The agent breaks the work into branches and can tackle parts in parallel.
  • 2X Deep Thinking: The most thorough mode, where the agent can hand parts of the job to specialised sub-agents.
Higher modes are more thorough but slower, so pick the lightest one that gets the job done.

What “tools” mean

A tool is a specific capability the agent can reach for while answering — for example running a query against your data, building a chart, or searching the web. Each agent comes set up with a sensible set of tools for each mode.
  • Some essential, behind-the-scenes tools always run and are never shown.
  • You can only enable tools your organisation’s admins have already approved for the agent.
  • Unticking every optional tool means the agent works with only the always-on essentials.
  • If an agent’s mode has no optional tools set up, the popover says so.

The Sources / Attachments controls

  • Attach file (paperclip in the composer): Opens the chat’s Attachments panel to add files to this conversation.
  • Data Sources (the Sources control at the top of the chat): Lets you pick which connected data sources this chat can read. Some data-focused tools and agent options only become available once at least one data source is attached.

Who can use it

  • Everyone who can edit a chat (owner, editor): Can type and send messages, stop a running reply, switch the Deep Thinking level, toggle web search, pick from the available agents, adjust tools for a turn, and attach files or data sources.
  • Viewers: Can read the conversation but cannot send messages or change these settings.
  • Admins: Decide which agents and which tools are available in the first place.