Agents

You talk. It makes.

Who you talk to. It finds a direction, writes the prompt, and refines the idea — against the canvas, not an empty chat.

Voice in, canvas out.

What an agent does.

Direction

Find a way into the campaign. Winter. These products. Doran. The agent writes the prompt against what is already on the board.

Direction still for a winter product campaign

Against the canvas

The work lands where you can see it. Empty chat is not the product.

Campaign still sitting on the canvas

Refine

Talk again after you see the board. Not after the context is gone.

Canvas board with the brief still in view

Empty chat is not the product.

The agent is who you talk to. Taste is Memory. The still belongs on the canvas.

Who you talk to

Direction, the prompt, the next pass. Not a model picker.

Taste is Memory

Agents do not have taste. Taste is Memory applied.

You correct it

Watch the process, own it, and fix what is off on the canvas.

Questions we get.

Frequently asked questions

Do agents have taste?

No. Agents do not have taste. Taste is Memory applied. The agent is who you talk to.

Is this a chat?

Chat gives you one picture, then the context is gone. The agent works against the canvas, so the next still belongs to the same campaign.

What do I talk about?

The campaign. Winter. These products. Doran. You should not have to become a prompt expert.

Which agents work with Gavana?

The ones you already use — Claude, ChatGPT, Gemini, Cursor, Codex, Grok, and others. They work against the canvas, not an empty chat.

Say the campaign.