On October 7, 2026, Search Engine Journal reported that WordPress has unfurled a new plugin, known as the canonical MCP Adapter, which facilitates developers in bridging AI systems with WordPress websites through a singular, streamlined plugin.
While the plugin itself is not a novel concept, its new location is significant. It has now been integrated into the official WordPress.org plugin directory, enabling any user to effortlessly install it from their dashboard.
Understanding MCP
MCP, or Model Context Protocol, is characterized by Search Engine Journal as an open standard developed by Anthropic, designed to ensure secure connectivity between AI systems and various applications, data, plugins, tools, and websites. This connectivity empowers AI to access information and execute specific tasks.
The WordPress.org page succinctly explains the protocol’s significance: without MCP, each AI client necessitates a distinctive integration method.
MCP, however, allows for capabilities to be listed in the form of tools, resources, and prompts, accessible to any MCP client.
Within the realm of a WordPress site, this translates to the ability for AI to interact with various facets of WordPress, ranging from minor actions, such as publishing a post, to larger functions, like manipulating design elements.
New Developments
Although the adapter had previously been available on its GitHub releases page, its presence in the official WordPress.org repository marks a significant milestone, with over 40,000 active installations reported.
Version 0.7.0 is the inaugural iteration to be formally published in this repository, with Search Engine Journal referencing it as version 0.70, while WordPress.org lists it as 0.7.0.
Jason Adams, Automattic’s Director of Engineering, AI, declared the release on X, dubbing it the canonical MCP plugin for WordPress.
He emphasized that it was meticulously designed for backward compatibility with earlier MCP versions and that it utilized the Abilities API to prevent duplicated functions in authorization.
Importance of Backward Compatibility
A commenter on X (formerly Twitter) remarked on the longevity of WordPress plugins, which indeed underscores the relevance of compatibility.
Adams concurred, clarifying that while the team cannot influence the backward compatibility of the MCP protocol itself, a pathway within the WordPress ecosystem was essential.
To elaborate, the MCP is an evolving standard, and the plugin functions as a stabilizing mechanism for those unable to update their software at the same pace.
Functionality of the Adapter
The plugin’s page on WordPress.org distinctly states that this is not an AI-inference plugin. The MCP Adapter serves as the server and transport layer; it does not possess AI capabilities of its own.
Instead, its primary function is to reveal abilities—individual capabilities registered by the core system or another plugin, such as reading posts or modifying a setting.
The adapter gathers the available abilities on a specific site, rendering them accessible to MCP clients and managing protocol-level negotiations, permissions, and error-checking.
This leads to a pivotal detail: for the plugin to operate effectively, a site must have its abilities adequately exposed, which can be done via the core system or by another plugin (or less commonly, a theme).
Notably, the plugin is a component of the AI Building Blocks for WordPress initiative.
Notable Features
The plugin page enumerates several key features:
- Ability-to-MCP conversion. Converts WordPress abilities into MCP tools, resources, and prompts.
- Multi-server management. Developers can establish multiple MCP servers, each with unique settings.
- Extensible transports. Built-in support for HTTP and STDIO, with options for custom configurations.
- Error handling. Default behavior aligns with WordPress logging, accommodating custom handlers.
- Observability. Metrics tracking with adjustable handlers.
- Permission control. Granular checks for all exposed plugin functionalities.
AI Agent Connectivity
The adapter registers MCP endpoints via its inherent HTTP transport. An MCP-compatible client establishes a connection to the endpoint, logging in as a WordPress user. The plugin accommodates two specific MCP revisions, 2025-11-25 and 2026-07-28.
Security and Control Considerations
Granting AI access to a website inevitably raises security concerns. The plugin’s page addresses several of these queries.
Access is strictly opt-in. Merely having a registered ability does not permit automatic usage; explicit marking as public is required for an ability to be accessible via MCP.
However, permission checks are performed consistently, dependent on the currently logged-in user, safeguarding against unauthorized method access.
A prudent strategy would advocate for exposing only those capabilities essential for AI agents while utilizing restricted accounts whenever feasible. This suggestion, however, is my own rather than articulated on the plugin page.
Security vulnerabilities should be reported according to the guidelines established in the project’s SECURITY.md file, and not disclosed publicly.
Installation Prerequisites
The plugin mandates WordPress version 6.9 or higher and PHP version 7.4 or beyond. It has been verified up to WordPress 7.1.3.
To install, navigate to Plugins > Add New within your WordPress administration panel. Search for MCP Adapter, click “Install Now,” activate the plugin, and register the desired abilities for exposure, or alternatively, install a plugin that facilitates this.
For developers who prefer command-line operations, a WP-CLI command can be utilized to install and activate the plugin. Developers whose plugins depend on the adapter can include a Requires Plugins header within their code.
Changes in Version 0.7.0
The release of version 0.7.0 occurred on October 2, 2026, revealing several significant changes for developers.
This version supports only two MCP schema revisions. Clients attempting to utilize a schema older than what the server specifies will continue to function, but with the 2025-11-25 schema.
Various generated classes and validators have been eliminated in this release. Invalid fields in requests will no longer be silently removed or rectified.
Custom transports now require handling specific functionalities rather than serving as generic middlemen, and JSON-RPC batches are no longer permitted.
New features include support for the 2026-07-28 revision and a sessionless discovery phase. Tools that are registered for direct invocation can now solicit user input if required under the new revision. Additionally, a new context object is now available, which remains unaltered during the request process.
A crucial update is the deprecation of bundling the adapter within other plugins; developers must now install the MCP Adapter plugin as a separate entity. A migration guide delineates all necessary changes.
Previous Versions
Version 0.6.0, released on August 12, 2026, requires WordPress 6.9 and eliminates the necessity for a standalone Abilities API plugin installation.
This iteration also modifies how the default server marks abilities as public and resolves several issues related to resource and metadata handling.
A subsequent release, version 0.6.1, followed the next day to rectify a packaging error that could result in fatal errors on certain sites. No changes to the API, hook, or protocol levels were made in this version.
A Collaborative Endeavor
The MCP Adapter plugin is described on its page as a community-driven project. Contributors include individuals from WordPress.org, Ovidiu Galatan, and David Levine.
Bug reports and feature requests can be submitted through the GitHub repository, while support is accessible via a dedicated forum on WordPress.org.
Implications for Site Owners
For the majority of site owners, the landscape remains unchanged as of today. The adapter serves as a resource for developers and plugins intending to collaborate with AI agents.
The advantages for developers are apparent, as there is now a singular, official plugin upon which to build, mitigating the risk of competing versions, while development on older bundled methodologies is anticipated to cease shortly.
Site owners wishing to leverage AI tools for content management or configuration should bide their time for plugins that register their capabilities, carefully scrutinizing what each one offers before enabling the functionality.
Q 1: What is the WordPress MCP Adapter?
It is an official plugin designed to connect WordPress to AI agents via the Model Context Protocol, exposing WordPress functionalities as MCP tools, resources, and prompts.
Q 2: What is MCP?
Model Context Protocol is an open standard intended for the integration of AI systems with applications, data, and tools, developed by Anthropic according to Search Engine Journal.
Q 3: What is new in this release?
The 0.7.0 version is now available in the official WordPress.org plugin directory, a shift from its prior GitHub exclusivity.
Q 4: How many sites are utilizing it?
According to the plugin page, it boasts over 40,000 active installations.
Q 5: Does the plugin introduce AI features to my site?
No, the WordPress MCP Adapter functions solely as the server and transport layer—AI functionalities necessitate the installation and configuration of the requisite abilities from either the core system or additional plugins.
Q 6: Does it expose all functionalities automatically?
No, capabilities are exposed selectively; first, they must be marked as public, and user permissions must be verified.
Q 7: What are the requirements for running it?
A minimum of WordPress 6.9 and PHP 7.4 or higher is required. It has been tested up to WordPress 7.1.3.
Q 8: How do I install the plugin?
Navigate to Plugins, then Add New in your WordPress administration panel, search for MCP Adapter, and proceed with the installation. A WP-CLI command can also be employed.
Q 9: What modifications occurred for developers in version 0.7.0?
This version serves two MCP revisions, removes several classes and validators, rejects JSON-RPC batch requests and disallows bundling the adapter as a library.
Q 10: Where should I report any bugs?
Bugs should be reported within the GitHub repository. For security concerns, please refer to the advisory within the project’s SECURITY.md file.
Source link: Almcorp.com.
