The dock ships with three personas. Each one has a different system prompt, a different set of allowed tools, and a different UI body. Pick the one that matches your task.
Chat Agent
Reads one or more reference images and proposes an enriched prompt for image or video generation. Use it when:
- You have a strong visual reference but only a vague mental description.
- You want to learn what details a good prompt actually includes.
This persona does not generate images itself — it only writes prompts. Copy the suggested prompt into the Generate Image panel.
Board Agent
The multi-step actor. The Board Agent can:
- Read the board — list elements, summarise content, find specific items.
- Place media — generate or import, then position on the canvas.
- Group and organise — create frames, apply auto-layout, run Smart Grouping.
- Build video projects — assemble a timeline in the Video Editor.
- Edit HTML cards — rewrite an HTML card’s contents in place (the
update_element_htmlaction). - Build and render video projects — including native MotionFast timeline layers.
- Search — run Visual Search across the SerpAPI integration.
Every proposed action shows up as a card with Apply / Reject. Approve them one by one, or click Apply all to commit a sequence.
The Board Agent is the most powerful persona and the most expensive in tokens — multi-step plans cost more than a single chat turn.
The Board Agent is a Pro-only feature.
Switching mid-conversation
The agent dropdown can be changed between messages. Doing so resets the conversation context — the new persona does not see the previous transcript. Make sure to finish or copy any in-progress work before switching.
Leave a Reply