Effectively Using the Creative Agent

This lesson covers how to work effectively with the Creative Agent: how to prompt it well, make page-wide changes, pull content from your Library, add extra context, and more.

Next Lesson:📄 Creating Advanced Sections

Versioning

Mutiny automatically saves versions of your asset as you work. To go back to a previous version, click the Version History button in the top right of the editor, select a version, and confirm. Any changes made after that version will be lost.

If you want to undo only the agent's changes without affecting your manual edits, hover over any message in the Chat tab and click Restore This Version.

If the agent runs into errors, click the menu in the prompt box and select Clear Chat History to reset it. From the same menu, you can also switch which AI model is being used if a provider is experiencing an outage.

Prompting Best Practices

The agent understands natural language, so you can talk to it like a teammate. A few habits will get you better results.

Be specific about where you want changes made. Instead of "update the hero," try "rewrite the hero subheader to be two sentences focused on pipeline efficiency." If you want changes within a specific section, click that section first and then type your prompt. If you want a change across the entire page, open Chat without selecting anything, and the agent will apply it everywhere relevant.

You can also use the agent for non-design tasks, like asking it to suggest personalization ideas before committing to any edits.

Multi-Element Changes

To make a change across every instance of an element on the page, open the Chat tab without selecting anything and describe what you want globally. For example: "Change the CTA text for all buttons on the page to Get a demo." The agent will scan every section, find every matching element, and update them. After it runs, scroll through the page to confirm the changes look right.

You can also ask the agent to proofread the asset, check for consistency with a specific topic, or optimize the page for mobile.

Pulling From the Library

If your Library is populated with content, the agent will draw from it during asset creation and editing. For example: "Create a new section below the hero and add the three most relevant case studies." The agent will select the appropriate resources, build the section, and populate it with the correct links.

If you want to swap out an image the agent used, hover over it and click Replace to choose an image from the Library, upload a new one, or crop what is already there.

Uploading Context

The plus button in the prompt box lets you attach context directly to your message before sending it. You can upload a screenshot of a design you want to replicate, a call transcript, a messaging document, or any other reference file. The agent will use that material as direction when executing your request.

Using the Mic

Click the mic icon in the prompt box to dictate your prompt instead of typing it. Everything you say is transcribed and sent to the agent the same way a typed prompt would be.