The Bricks 2.4 Beta ships a feature that changes what page builders are capable of: a native Model Context Protocol connection that lets external AI clients build directly inside your Bricks canvas. We have been testing it on staging environments, and the results are more nuanced than the launch excitement suggests. Here is what it actually does, how to set it up, and where it genuinely belongs in your workflow.
For context on how Bricks MCP compares to standalone WordPress AI agent tools, read our guide on Novamira MCP: AI-Powered WordPress Features and our comparison of top Novamira alternatives for AI-powered WordPress development.
TL;DR: Bricks 2.4 Beta AI MCP
- Bricks 2.4 Beta introduces a native MCP server that lets AI clients like Claude and Codeex build directly inside the builder.
- The AI reads your database, CSS variables, breakpoints, and global classes before generating anything.
- A single responsive hero section took 13 minutes to generate in real-world testing. That is slower than manual template customisation.
- The real value sits in backend tasks: complex ACF relationships, database queries, and dynamic content filters.
- Test on staging only. Never run the Bricks MCP server on a live production site.
What the Bricks 2.4 Beta AI MCP Connection Actually Is
Most AI integrations in page builders work the same way. You type a prompt, the AI returns CSS or HTML, and you paste it somewhere. Bricks 2.4 Beta does something genuinely different. Its native MCP server lets an external AI client connect directly to the builder and build inside it using real Bricks elements rather than raw code output.
The AI reads your WordPress database before generating anything. It accesses your active breakpoints, existing CSS variables, and global classes. It builds using native Bricks container elements and applies BEM class naming conventions consistently. This is not a code suggestion tool. It is an agent operating inside your builder environment with awareness of your specific site structure.
This approach puts Bricks 2.4 Beta in the same architectural category as tools like Novamira, which provides AI agents with PHP-level WordPress access. The difference is that Bricks builds this connection natively into the page builder rather than requiring a separate plugin. Compatible AI clients at launch include Codeex and Claude.
âšī¸ Note: Bricks 2.4 is a beta release. All testing should happen on a staging or local site. Running the MCP server on a live production website carries real risk of unintended changes to your database and file structure.
Want to see the setup in action? Watch the full first-look walkthrough below to see how Bricks 2.4 Beta handles live AI design, step-by-step:
How to Set Up the Bricks 2.4 Beta MCP Server
The setup process takes under five minutes and requires no coding knowledge.
Step 1: Enable Bricks Abilities
Go to your WordPress dashboard, click Bricks, and open the new AI settings tab. Flip the master switch to enable the builder’s capabilities. This activates the MCP server infrastructure within your installation.
Step 2: Generate Connection Credentials
Under the Connect AI Client to MCP section, create a name for the connection, generate a secure password, and copy the provided connection string. Keep this string safe. It is the authentication key your external AI client uses to connect.
Step 3: Connect Your AI Client
Navigate to your external AI tool, whether that is Codeex, Claude, or another compatible client, and paste the connection string to authenticate the bridge between your AI client and the Bricks builder.
Step 4: Load and Verify Skills
Return to the Bricks settings, copy the skills prompt, and use it to pull the official Bricks skills from GitHub into your AI client. Run the verification prompt to confirm all 44 standard capabilities are active and greenlit. The skills define what actions the AI agent can take inside the builder, from creating containers to applying responsive breakpoints.
đĄ Pro Tip: Run the full verification check before starting any build task. If any of the 44 capabilities fail verification, the AI may partially complete tasks or produce inconsistent output. Catching failures at the skills stage costs seconds. Catching them halfway through a complex build costs significantly more.
Real-World Test Results: What Happened When We Used It
We gave the Bricks 2.4 Beta MCP connection a straightforward test: build a fully responsive hero section from a wireframe screenshot. The AI read our CSS variables, applied BEM naming correctly, and used native Bricks container elements throughout. The layout accurately matched the wireframe.
The time it took was the problem. Thirteen minutes elapsed from prompt to completed hero section. A professional template library import and manual customisation of the same section takes under two minutes. For visual layout work, the AI MCP connection cannot yet compete on speed with established template-based workflows.
We also noticed the AI occasionally placed styling values into custom CSS fields rather than using Bricks’ native visual styling panels. The output was technically correct but created additional cleanup work to align with how experienced Bricks developers prefer to manage styles.
The honest conclusion from visual layout testing: this tool is not ready to replace your template library for standard design work. That is not what it is for.
Where the Bricks AI MCP Connection Actually Belongs
The meaningful use cases for Bricks 2.4 Beta AI sit firmly in backend complexity rather than visual design generation.
Complex ACF relationships are where the AI delivers clear value. Setting up bidirectional relationships between custom post types, configuring conditional field display rules, and mapping ACF field groups to dynamic content templates are all tasks the AI handles faster and more accurately than most developers manage manually.
Database queries are another strong use case. Writing complex WP_Query arguments, building meta query structures, and debugging query logic that returns unexpected results all benefit from having an AI agent with direct database access and full awareness of your site’s existing post types and taxonomy structure.
Dynamic content filter setup rounds out the current practical value proposition. Configuring repeater fields, setting up dynamic data connections between Bricks and ACF, and building template conditions that respond to post type or taxonomy all suit the AI’s strengths better than responsive hero sections do.
Think of the Bricks AI MCP connection as a backend automation layer, not a design tool. Use your template library for visual layouts. Use the AI for the complex logic underneath those layouts.
How This Compares to Standalone WordPress MCP Tools
Bricks 2.4 Beta’s native MCP integration is interesting precisely because it narrows the scope of what the AI can access. Tools like Novamira grant AI agents full PHP runtime access to the entire WordPress installation. Bricks scopes access specifically to the builder environment and the site data relevant to building within it.
That scoping is both the feature and the limitation. It makes the Bricks MCP connection safer and more predictable for builder-specific tasks. It also means the AI cannot perform the broader operational tasks, bulk post updates, PHP debugging, or plugin configuration work, that full PHP runtime tools handle.
For teams whose primary AI workflow sits inside the Bricks builder, the native integration is the simpler path. For teams needing broader WordPress automation alongside builder work, Novamira or the alternatives in our Novamira alternatives guide remain the more complete solution. The two approaches complement each other rather than compete.
Additional Resources
- Novamira MCP: AI-Powered WordPress Features Guide
- Top Novamira Alternatives for AI-Powered WordPress Development in 2026
- Bricks: Visual Site Builder for WordPress
- Novamira: Full WordPress Access for AI Agents
We are genuinely excited about where Bricks 2.4 Beta points. Native MCP integration inside a page builder is a meaningful architectural step, and the fact that the AI reads your actual site data before generating anything puts it ahead of most AI builder integrations. The visual speed limitations are real and worth acknowledging honestly. The backend use cases, ACF setup, complex queries, dynamic content, are where this tool earns its place in a professional WordPress workflow right now. Test it on staging, find your workflow sweet spot, and watch this feature mature through the Beta cycle.
