Conversations
Each chat is a separate conversation, and your history is saved so you can come back to any of them.- New chat (⌘+J) starts a fresh conversation.
- Chats opens your history. Search it with Search chats…, and pin, rename, or delete any conversation. Pinned chats stay at the top, and a chat the agent is currently working in shows a live dot.
- Above the message box, Recent chat takes you straight back to your most recent conversation. While the agent is still working in one, it reads Working chat instead.
- An open chat’s menu offers Rename, Download chat (.md), and Delete.
Writing a message
- Type your request and press Enter. While the message box is empty, the microphone button lets you dictate instead.
- While the agent is working, the send button becomes Stop — click it to interrupt.
- Start a message with / to pick one of your skills for that request.
- Write an entry’s name in double brackets —
[[Acme Rebrand]]— to point the agent at that entry. - Use the + button to attach files. See Giving the agent context.
Choosing a model
The pill next to + in the message box picks the AI model. It lists every model your workspace can use, grouped by provider, with the workspace default first. Your choice applies from the next message you send and is remembered for next time. At the bottom of the same menu, the Thinking slider sets how much the agent reasons before it answers — Off, Low, Medium, High, or Extra High. Higher levels help with harder requests but take longer. Some models can’t read images. While images are attached, those models are unavailable in the menu. Which models are available depends on your workspace’s AI plan. If the workspace runs one fixed model, the pill is locked and says Model is set by Studiograph Cloud.What the agent can do
Ask in plain language; the agent picks the right action. Find and read- Search and read anything you can access — entries, folders, and everything linked to them.
- Follow connections: “what’s related to this project?”
- Read files in your library, including PDFs page by page.
- Look through an entry’s history and bring back an earlier version or a deleted entry.
- Create entries, edit them, rename them (links update automatically), move them between folders, and delete them.
- Build and edit artifacts — presentations, artboards, apps, and publications — frame by frame.
- Work with datasets: look up answers, add rows, and update cells.
- Generate images — “make a hero image for this case study”. Image generation is part of your AI plan.
- Import files into your workspace, save images from a web page, and attach assets from your library to entries.
- Search the web and read web pages. Web search is on by default; how it runs depends on your AI plan.
- Use your connected apps through connectors.
- Create, rename, and archive folders.
- Follow your saved skills for repeatable tasks.
- Run the onboarding interview to learn about you and your team.
The agent acts as you. It can only see and change the folders you have access to — nothing more.
Long conversations
In a long chat, the agent summarizes older messages to stay focused — you’ll see a Compacted divider where that happened. If a conversation grows too large for the model, the agent asks you to start a new chat.When the AI isn’t available
If the agent can’t reach its model — for example, because the workspace has run out of AI credits — the chat says so and offers a way to fix it. See Your AI plan.Next steps
- Giving the agent context — attachments, selections, and how the agent knows what you’re looking at.
- Skills — teach the agent your repeatable tasks.