AI Chat

AI Chat blocks let you have conversations with leading AI models right on your canvas. Ask questions, brainstorm ideas, generate content, and branch conversations to explore different directions — all without leaving your workspace.

Available models

Choose from three AI models using the selector at the top of the chat block:

You can switch models mid-conversation. The next response will use the newly selected model.

Starting a conversation

Create an AI Chat block from the block selector. Type your message in the input field and press Enter to send. The AI's response streams in real time — you can see words appear as they're generated.

Streaming & stopping

Responses stream in real time. If you want to stop a response mid-generation, click the Stop button. The partial response is kept in the conversation.

Branching conversations

Select text in an AI response to reveal the Branch button. Clicking it creates a new AI Chat block connected to the parent with a visual edge. The new block starts with the selected text as quoted context, letting you explore a tangent without losing the original conversation thread.

Custom titles

Conversations get an auto-generated title based on the first message. Click the title to rename it — useful for organizing multiple chat blocks on your canvas.

Markdown responses

AI responses are rendered with full Markdown support including headings, bold/italic text, lists, code blocks with syntax highlighting, and links.

Canvas AI panel

In addition to AI Chat blocks, every canvas page has an AI sidebar panel — click the sparkle button or press / to open it. The panel supports all the same models plus canvas tools: create, edit, reposition, resize, and delete blocks through natural language. See the Cobloc AI article for the full workspace-level AI experience.

Block connections

Connect image, note, todo, or document blocks to an AI Chat block by dragging from the connector dot on the content block to the AI Chat block. The AI automatically sees the connected content as context — including images via vision models and PDF text extraction. Up to 5 blocks can be connected. See Canvas AI for details.