Use Builder Agent in Builder Content
Builder Agent can assistant in the Builder Content UI. You describe a task in plain language, then the agent then creates pages, scaffolds sites, imports records, or updates content for you.
You review and approve the agent's work. You can then fine-tune individual entries in the Visual Editor. The agent follows your roles and permissions, so it can't do anything you couldn't do yourself.
Open and close the Builder Agent
The Builder Agent appears as a sidebar on the right side of the Content and Content Models pages. To make it appear, click the Open Content Agent button.
Start a chat
You can start a chat in two ways:
- Describe what you want in the chat input.
- Type
/to choose a built-in skill. Each skill has a short description.
A skill is a built-in workflow for a common task. The Content Agent includes three built-in skills:
| Skill | What it does |
Create page | Creates a new page for a page-type model. |
Scaffold site | Builds a set of pages from a sitemap, CSV, or site URL. |
Import records from CSV | Imports a CSV file into a model. The agent creates the model and fields if needed. |
You aren't limited to skills. The agent can do anything the Builder Content MCP server supports, including targeting and scheduling.
In the video below, the Builder Agent is prompted to create a new content entry.
Scaffold a site
Use the Scaffold site skill to create the pages for a new site. You can start from a sitemap, a CSV, or an existing site URL.
To scaffold a site:
- In the Content Agent, type
/and select Scaffold site. - Enter a site URL, or attach a sitemap or CSV.
- Review the structure the agent proposes. It lists the pages it plans to create and the pages it plans to exclude.
- Tell the agent what to change. For example, say “trim this to essential pages” or “add the pricing pages back.” Repeat until the list looks right.
- Choose to implement the list.
- Approve each batch of pages. The agent asks for approval five pages at a time.
The agent creates the pages as empty drafts and reports the URL path for each one. You can then add content to the pages from the agent or in the Visual Editor.
Import records from a CSV
Use the Import records from CSV skill to bring in many entries at once. For example, you can import blog posts exported from another CMS.
To import records from a CSV:
- In the Content Agent, type
/and select Import records from CSV. - Drag your CSV file into the chat.
- If your Space has no matching model, the agent asks where to import. Choose an existing model, or have the agent create one. You can rename the new model.
- Review the fields the agent proposes based on your CSV's columns and rows.
- Approve the request to create the model and import the records.
To see exactly what the agent sends, expand the request details before you approve.
The agent saves the model it creates. Later CSVs with the same structure import into it without setup.
Attach files and designs
You can attach context to any chat. You have three options:
- Import a Figma design to build a page from it.
- Pick an existing file from your Asset Library.
- Upload a new file.
For example, attach a design and ask the agent to build out your Contact Us page.
Approve the agent's actions
The agent asks for your approval before it changes your content. Each request shows what the agent will do and which entries it affects. Choose Approve to continue, or decline to stop.
You can choose how much approval the agent needs:
- Always ask (default): The agent asks before each change.
- Full access: The agent can create content and models without asking.
In either mode, the agent creates content as a draft. To publish, go to the entry on the Content page. You can use the usual content tools there to review and undo changes.
Review your chat history
The Builder Agent keeps a history of your chats. To see past work, open the chat history. Then select a chat to review what the agent did or to continue it. Chats are private to you.
Choose a model and give feedback
The chat includes two controls for tuning your experience:
- Use the model selector in the chat input to choose which AI model powers your chat. Each response also shows the credits used.
- Use the feedback controls on the agent's responses to share what worked and what didn't.
Connect other tools
You can add other MCP servers to the agent to bring in outside context. Examples include Google Docs, Notion, Jira, and Linear.
