WordPress 7 and the MCP Revolution: How AI Agents Are Reshaping the CMS That Powers 40% of the Web

On April 9, 2026, WordPress will ship the most significant update in its two-decade history. WordPress 7.0, the release that finally delivers Phase 3 of the Gutenberg project, is not just a collaboration update or an editor refresh. It is a platform-level transformation that embeds artificial intelligence into the core architecture of the CMS that powers more than 40 percent of the entire web. For organizations running headless WordPress with modern front-ends, and for agencies like The Berchtold Group that build and maintain those architectures, this release changes the game.

The headline feature is not a new block or a redesigned admin panel, though both are present. The headline is that WordPress 7.0 ships with a native AI Client, an Abilities API, a Connectors UI for managing AI provider credentials, and an official Model Context Protocol adapter that lets AI agents discover and execute WordPress functionality programmatically. That last piece, MCP, is the one that matters most for the future of content operations, and it is the one most organizations are not yet prepared for.

What Is the Model Context Protocol and Why Does It Matter

The Model Context Protocol is an open standard created by Anthropic and adopted across the AI industry. Think of it as a universal adapter, the USB-C of AI integrations. Before MCP, every connection between an AI tool and an external system required custom code. Connecting Claude to your CMS was a different integration than connecting it to your analytics platform, which was different again from connecting it to your project management tool. The result was an N-times-M integration problem that scaled poorly and fragmented the ecosystem.

MCP solves this by defining a single protocol with three core primitives. Tools are executable functions the AI calls to perform actions, like publishing a post or updating a menu. Resources are passive data sources the AI reads for context, like site analytics, taxonomy structures, or content inventories. Prompts are pre-configured templates that guide specific workflows, like a content brief or a publishing checklist. An AI agent that speaks MCP can connect to any MCP server and immediately discover what that server can do, what data it exposes, and what workflows it supports, without any custom integration code.

What WordPress 7.0 Actually Ships

The AI infrastructure in WordPress 7.0 is built on three layers that work together. Understanding each layer matters because they represent a fundamental shift in how WordPress thinks about extensibility, moving from hooks and filters designed for PHP developers to structured capabilities designed for both humans and machines.

The Abilities API: WordPress as a Composable Platform

The Abilities API, first introduced in WordPress 6.9, lets plugins register discrete units of work as standardized, discoverable, typed, and executable abilities. Rather than scattering functionality across custom REST endpoints and fragmented hooks, developers register each capability once. The API automatically handles data validation, access control, and discoverability. Every registered ability requires a permission callback before execution, so WordPress validates whether the acting user, human or AI agent, is authorized to perform the action.

This transforms WordPress into a composable platform where each feature becomes a reusable building block. A plugin that registers a "create-redirect" ability makes that capability available not just to the admin UI, but to any AI agent that connects to the site. The ability is the same regardless of who or what invokes it.

The WP AI Client: WordPress Calls AI

The WP AI Client is a fluent PHP API that lets plugin developers make AI requests from within WordPress. The core function, wp_ai_client_prompt(), supports method chaining for setting system instructions, attaching files, specifying model preferences with fallbacks, requesting structured JSON responses, and generating text. A JavaScript counterpart, wp.aiClient, mirrors the server-side functionality for use in the block editor.

AI providers ship as separate WordPress.org plugins. At launch, official provider plugins exist for Anthropic Claude, OpenAI, and Google Gemini, with community-contributed providers for Grok and OpenRouter already appearing. The new Connectors UI, located at Settings > Connectors in the admin sidebar, gives administrators a centralized interface for managing API credentials. Enter your API key, WordPress validates it server-side, and the credential is stored with proper masking. The system is entirely optional. WordPress 7.0 works exactly like previous versions without any API keys configured, adding zero overhead unless you actively make AI calls.

The MCP Adapter: AI Calls WordPress

The MCP Adapter is the piece that completes the circle. While the WP AI Client lets WordPress call AI, the MCP Adapter lets AI call WordPress. It implements the Model Context Protocol in the scope of a WordPress site, adapting every ability registered through the Abilities API into MCP tools, resources, and prompts that AI agents like Claude Desktop, Claude Code, Cursor, and VS Code can discover and execute.

The security model is role-based and granular. AI agents operate as authenticated users, and their capabilities are limited by assigned roles. An agent running under an editor account cannot delete plugins. An agent without publishing rights cannot push content live. At no point is data shared between the MCP server and the LLM without the user's complete control, and data retrieved via MCP tools is used only for the original request, never for model training.

Beyond AI: What Else Ships in WordPress 7.0

The AI infrastructure is the strategic headline, but WordPress 7.0 is a deep release across the board. Phase 3 of Gutenberg delivers real-time collaboration, letting multiple users edit the same post or page simultaneously with data syncing, stabilized notes, and offline editing support. A new HTTP polling sync provider ships by default, with options for plugins or hosts to enable WebSocket support for lower-latency collaboration.

DataViews replaces the legacy WP List Tables with a modern, app-like interface for managing content, users, and other admin data. The editor canvas moves toward full iframing, creating a sandboxed environment that makes WYSIWYG significantly more accurate. New blocks include video embed backgrounds in the Cover block, a responsive Grid block, and new Icons and Breadcrumbs blocks. Client-side media processing handles image resizing and compression in the browser rather than on the server, reducing hosting load.

What This Means for Headless WordPress Architectures

For organizations running headless WordPress, where WordPress serves as the content back-end and a modern JavaScript framework renders the front-end, WordPress 7.0 is a force multiplier. The expanded REST API introduces new endpoints for advanced site analytics, user management, and content scheduling that were previously unavailable through the API. Built-in response compression and conditional GET support improve REST performance. The Abilities API provides granular, capability-based access control that replaces older role-based permission checks, simplifying headless authentication flows.

WPGraphQL 2.x remains the preferred query layer for most headless builds, with automatic persisted queries, cache control directives, federation support, and built-in dataloader batching that reduces database queries by 60 to 80 percent. Combined with Next.js on the front-end, which brings stable Server Components, improved streaming SSR, and Turbopack as the default bundler, the headless WordPress stack in 2026 is faster, more capable, and more developer-friendly than it has ever been.

The MCP layer adds a new dimension. In a headless architecture, content authors work in WordPress while developers build the front-end in a framework like Next.js. The MCP Adapter means AI agents can now operate in that same content layer, creating drafts, running content audits, checking SEO metadata, updating taxonomies, all through structured, permission-controlled interactions that respect the same roles and workflows human authors use.

How The Berchtold Group Is Already Building for This Future

At The Berchtold Group, we have been building at the intersection of headless CMS architecture and AI-native tooling since before WordPress shipped any of this infrastructure. Our clients get the best of both worlds: the content authoring familiarity of WordPress, which remains the platform that more than 40 percent of the web runs on, combined with the performance, developer experience, and best-practice standards of front-ends built with Next.js, Tailwind CSS, TypeScript, and React.

That headless architecture is not just a technical preference. It is a strategic position. When WordPress ships native MCP support, organizations with decoupled architectures are the ones best positioned to take advantage of it. The content layer becomes an AI-accessible service, the front-end remains fast and framework-native, and the entire stack is composable in ways that monolithic WordPress installations cannot match.

We have already built the MCP infrastructure our clients need. The Berchtold Brand MCP is a production Model Context Protocol server that gives AI agents structured access to our clients' brand assets, voice guidelines, content standards, and design tokens. When a content author uses Claude or another AI assistant, the Brand MCP ensures that every piece of generated content is grounded in the client's actual brand system, not generic AI output. It is the brand guide as content operating system made real, a machine-readable layer that governs how AI represents your organization across every touchpoint.

When WordPress 7.0 ships on April 9 with its own MCP Adapter, our clients will have a connected ecosystem. WordPress exposes content abilities via MCP. The Brand MCP exposes brand governance via MCP. AI agents can interact with both through the same protocol, creating content that is not only technically correct in the CMS but brand-aligned from the first draft. This is not a future roadmap item. It is infrastructure that exists today, waiting for WordPress to catch up.

The Shift from AI-Compatible to AI-Native

The WordPress community has experimented with AI plugins for years. Jetpack AI, Yoast SEO's AI features, various content generation tools. These are bolt-on solutions that treat AI as an add-on to an existing workflow. WordPress 7.0 represents something categorically different. By embedding the Abilities API, the WP AI Client, and the MCP Adapter into core, WordPress is making a platform-level bet that AI agents are not a feature but a first-class interface to the CMS.

This shift from AI-compatible to AI-native matters because it changes who builds on the platform and what they can build. Plugin developers can register abilities once and have them work for human users and AI agents simultaneously. Hosting providers can pre-configure AI credentials via environment variables, giving every site on their platform instant access to AI capabilities. Content teams can delegate routine tasks to agents that operate within the same permission structure as human editors.

The organizations that benefit most from this shift are the ones that have already separated concerns, that have already decoupled their content layer from their presentation layer, that have already invested in structured brand governance that machines can consume. In other words, the organizations that are already running headless WordPress with a modern front-end and a brand system designed for the AI era.

What You Should Do Before April 9

• •Audit your WordPress installation for PHP 7.4 minimum compatibility. WordPress 7.0 requires PHP 7.4 at minimum and recommends PHP 8.3 or later for best performance. MySQL 8.0 or MariaDB 10.6 are now required.
• •Evaluate your headless architecture. If you are still running monolithic WordPress, consider whether a decoupled approach with a Next.js front-end would better position you for the AI-native features WordPress 7.0 enables.
• •Review your brand governance. The value of MCP-connected AI agents depends entirely on the quality of the brand system they draw from. If your brand guide is a PDF on a shared drive, it cannot serve as an input to machines. If it is structured, tokenized, and machine-readable, it becomes the operating system for every AI interaction.
• •Test WordPress 7.0 Beta 2 in a staging environment. Beta 2 shipped February 26 with the Connectors UI and more than 70 fixes since Beta 1. Test your plugins and themes now, not on release day.
• •Talk to your agency about MCP readiness. If your agency is not already building MCP infrastructure, they are behind. The organizations moving fastest are the ones whose partners anticipated this shift and built the tools before WordPress shipped the protocol.

The Bottom Line

WordPress 7.0 is not just another version bump. It is the release where the platform that powers 40 percent of the web becomes a first-class participant in the AI agent ecosystem. The Abilities API, the WP AI Client, the Connectors UI, and the MCP Adapter together create a two-way bridge between WordPress and every major AI model, all built on an open protocol that the entire industry is converging on.

At The Berchtold Group, we build headless WordPress architectures with Next.js, Tailwind CSS, TypeScript, and React because we believe the best digital experiences come from pairing the content authoring platform that content teams already know with the front-end technology that delivers the best performance and developer experience. WordPress 7.0 validates that architecture by making the content layer AI-accessible through MCP while leaving the presentation layer free to evolve independently.

Our clients are already using the Berchtold Brand MCP to ensure their AI-generated content is brand-aligned by default. When WordPress 7.0 ships on April 9, that existing MCP infrastructure connects directly to WordPress's new AI layer, creating a unified, brand-governed, AI-native content ecosystem that most organizations will spend the next year trying to build from scratch.

The future of content management is not WordPress or AI. It is WordPress and AI, connected through an open protocol, governed by your brand system, and delivered through a modern front-end that performs as well as it looks. That future is not coming. For our clients, it is already here.