Get the plugin
WP MCP/Integrations/Page builders

Page builders integration

Beaver Builder AI agent tool reference

Beaver Builder keeps a page as a node map in post meta: the published layout, the editor's working draft, their settings twins, and an enabled flag. WP MCP reads the node tree and writes either one operation addressed by node id or a whole tree, doing what Beaver Builder itself does when it publishes: the new map goes to the published layout and, when the page has a draft, to the draft too, so the editor opens on what is live, and the layout asset cache is cleared. A page whose draft holds unpublished editor changes is refused until that draft is published or discarded, so an agent never silently throws away someone's work. Every row is post meta, so the post snapshot restores it exactly.

detect-builderget-builder-contentupdate-builder-content

In practice

How Beaver Builder work goes with an agent.

Beaver Builder stores a page as a flat map of nodes, not a nested document. Every row, column group, column and module is its own entry keyed by node id, with a parent id and a position among its siblings. Those ids are stable, so WP MCP addresses nodes by id instead of by path, and an edit touches only the changed node plus the position of its siblings. Everything else in the map serializes exactly as before. Placement follows Beaver's own nesting rules, so a column cannot be dropped straight into a row.

Update the rich-text module in the intro row of the About page.

Runs get-builder-content for the node tree, then update-builder-content with operation update, the module's node id and attrs. The settings are merged, null removes a key, and the module keeps its settings.type slug. The published layout, and the draft when one exists, are written together.

Add a new row with one column and a text module at the top of the page.

Runs update-builder-content with operation add at the top level and index 0, with a row element that nests a column-group, a column and a module. New nodes get fresh ids, siblings are renumbered, and the layout asset cache is cleared afterwards.

Put a column directly inside the first row.

Refused before anything is written, with the message that a column cannot go inside a row. Columns belong in a column-group, which belongs in a row or another column, so the agent retries with the right nesting. No snapshot is taken for a refused call.

Replace the whole layout of the staging landing page with this JSON.

Runs update-builder-content with content set to a JSON array of nodes, which is converted back to the node map. The builder flag is enabled, the draft is kept in step, and the post and layout meta are snapshotted first so rollback-operation restores the old map.

What your AI can do

Real Beaver Builder operations, every write reversible.

  • Detect Beaver Builder pages
  • Read a page's node tree
  • Update, add, remove or move one node by id, or replace the whole tree
  • Keep the published layout and draft in step and clear the asset cache
  • Refuse pages with unpublished editor changes instead of overwriting them
  • Roll back any layout write exactly

Example prompts

Example Beaver Builder prompts.

Is the About page built with Beaver Builder?

Update the button text in the hero module on the Beaver Builder landing page.

Remove the empty column from the second row of /contact.

Under the hood

The Beaver Builder calls, under the hood.

wpmcp, beaver-builder

get-builder-content({ post_id: 44 }), node tree

update-builder-content({ post_id: 44, builder: "beaver-builder", operation: "update", path: "<node id>", attrs: {…} }), snapshotted, restorable

Safety and access

What can be undone on Beaver Builder, and what costs what.

Recoverable on Beaver Builder

  • The published layout, the draft copy and the enabled flag are all post meta, so one snapshot restores them together.
  • The per-layout asset cache is cleared after a rollback as well as after a write, so the page does not keep serving old CSS.

Beaver Builder limits and opt-ins

  • A page whose draft differs from its published layout is refused. The read reports draft_pending so an agent can see it before trying.
  • Beaver writes have no content hash. The draft check is the sole guard against a concurrent edit, so read and write in one pass.
  • The layout settings rows are never touched, and post_content keeps the text copy from the last Beaver publish, so a plugin that reads post_content sees the old text until then.
  • The design_system, templates and catalog scope reads are not available for Beaver Builder.

Free or Pro: All three builder tools are Pro for Beaver Builder: detect-builder, get-builder-content and update-builder-content, with no opt-in filter. On the free tier, export-content with mirror:true writes Beaver Builder pages to JSON files for git, and a page's meta, layout included, is copied by duplicate-post.

Part of WP MCP Pro

The Beaver Builder integration is part of WP MCP Pro.

One Pro licence unlocks the Beaver Builder tools and every other Pro integration on this site: Elementor deep editing, the other page builders, the forms adapter pack and more, 133 Pro tools in all. The free plugin still covers the rest of WordPress and the full snapshot and undo engine, so you can try the safety model before you buy.

Frequently asked

Beaver Builder MCP: questions, answered.

What happens if someone is mid-edit in Beaver Builder?

If the page's draft differs from its published layout, the write is refused. Writing over the draft would discard those changes and writing beside it would be overwritten on the next publish, so the draft has to be published or discarded first.

Is the rendered text copy updated?

post_content keeps the text copy Beaver Builder rendered on its last publish; rendering a new one needs the plugin's front end, so it is left alone. The layout itself is what the page renders from.

Why are nodes addressed by id and not by path?

Because Beaver Builder keeps stable node ids and a flat map, so an id points at the same node however the layout is rearranged. Other builders in the suite have no stored ids and use dotted paths instead.

Does an edit show up in the Beaver Builder editor right away?

Yes. When the page has a draft, the same node map is written to it, so the editor opens on what is live. A page with no draft keeps none; the editor copies the published layout into a new draft when it opens.

Does a write change the page's post status?

No. It writes the published layout row, the way a Beaver publish does, so a page that is already live shows the change at once. A draft or private post stays as it is. WP MCP does not publish Beaver pages.

Does it clear Beaver Builder's cached CSS?

Yes, when Beaver Builder is loaded: it deletes the layout's cached assets the way the builder's own publish does. If the plugin is not loaded, the files are simply rebuilt from the stored layout the next time it is.

Is there an AI agent for Beaver Builder?

Yes, through WP MCP rather than Beaver Builder itself. An agent you choose, such as Claude or Cursor, can detect Beaver Builder pages. WP MCP is a separate plugin, not a built-in AI feature of Beaver Builder. On the undo side: The published layout, the draft copy and the enabled flag are all post meta, so one snapshot restores them together.

Point your AI at Beaver Builder.