Skip to main content

Chat

Chat is the conversational surface for thinking through ideas, asking questions about your product, and getting AI-recommended next steps. It sits in a panel on the right side of every page, so you can talk to it while you're looking at analytics, a roadmap, code, or a launch draft — whatever you have on screen becomes context.

AI Expedite chat panel docked on the right side of the workspace

Sending a message​

Type into the "How can we help..." box and press Enter to send. Shift + Enter adds a new line.

When the box is empty, the send button switches to a microphone — click it to start a continuous voice conversation. The icon turns into a live equalizer while it's listening, and a voice-speed slider appears at the top of the panel so you can tune playback. Click the voice name next to the slider to cycle through TTS voices.

Attaching files​

Three ways to add an image, PDF, screenshot, or doc to a message:

  • Drag and drop anywhere onto the chat panel — a "Drop files here" overlay confirms the target.
  • Paste an image straight from the clipboard while the input is focused.
  • Pick a file with the attachment button next to the input.

Attachments preview as thumbnails above the input. Click a thumbnail to view full-screen, or hit the small × to remove one before you send.

Picking a model​

The model selector sits at the top of the chat panel. Click it to switch between Claude, Gemini, GPT, and any other models your workspace has enabled. The selection persists per chat — start a new chat to get a fresh default.

📷 Screenshot: Model selector at the top of the chat panel.

Pinning context with prompt paths​

Below the input, prompt path chips let you pin specific things as context for the next message — a document, a meeting transcript, or a saved workflow. Click a chip to attach it; click the × on a chip to detach.

The leftmost chip controls workflow mode:

  • Auto (default) — Chat decides which workflow to run based on what you're asking.
  • A specific workflow — locks chat into that workflow's steps for this conversation. Useful when you know exactly what you want (e.g. "Generate a launch post" or "Open a PR for this fix") and don't want chat to second-guess.

Click the chip to cycle through your workflows.

History and new chats​

  • + (top right of the panel) — start a new chat. The current one stays in your history.
  • ☰ — open the chat history drawer. Past chats are listed newest-first; click one to resume it.
  • The current chat's URL updates as you go, so you can copy the link to share or bookmark it.

What lives inside a chat​

A chat thread renders more than plain text. When the AI runs a tool — terminal commands, code searches, GitHub PRs, web searches, image or video generation, document edits — each tool call appears inline as its own message with a status, output, and (for terminal sessions) live streaming output. Click any tool message to expand or collapse its detail.

Each AI reply also shows a small model attribution badge so you can tell at a glance which model produced which answer when you switch mid-conversation.

Layout​

  • Resize — drag the left edge of the panel to make it wider or narrower (300–1000px).
  • Collapse — the » button collapses the panel into a floating chat bubble; click the bubble to expand.
  • Full-screen — open /<workspace>/chat directly to use chat as the entire viewport.

Tips​

  • Chat reads what's on screen. If you want it to focus elsewhere, navigate first, then ask.
  • For multi-step asks, picking a specific workflow up front is usually faster than letting Auto figure it out.
  • Voice mode is hands-free end-to-end — speech-to-text on the way in, text-to-speech on the way out — useful when you're walking through code or a draft on screen.