WordPress 7.1 Polished the Editor. It Also Let the Agents In.

When WordPress 7.1 launched on August 19 with a keynote at WordCamp U.S. in Phoenix, the applause was for everything you would expect: a cleaner editor, finally reliable rendering of the admin bar, and image editing that no longer stressed your server. Well done and long overdue.

But the change that will ultimately define Mary Lou — 7.1’s codename — as a milestone isn’t yet dramatic enough to yield an impressive screenshot. It’s instead an API largely unfamiliar to most site owners that controls what information your site is permitted to provide to an AI agent.

Even so, the release is enormous. The number of changes incorporated in the release candidate exceeded 145 compared with Beta 4, of which 88 were for core components and 57 for the editor.

The editing experience was completely isolated for all pages using an iframe. The admin bar was continuously visible on all screens and eliminated the need to search for it depending on which page had previously been edited. And all processing of media was performed entirely within the browser, which removes the need for server time and resources for image cropping, resizing, and conversion to formats such as HEIC, AVIF or WebP, according to reports by WPBeginner.

Other important features of the release included a new block for displaying playlists, shared links to various versions of the same content, inline editing of images, and notification of mentions by other users within notes. Considerable efforts also were directed toward achieving greater levels of accessibility, with resolution of 46 tickets related to that goal.

Overall, it was an outstanding release. However, none of the changes elicited the degree of excitement that existed among developers.

The major technical accomplishment of the release was continued development of the Abilities API to provide a complete, machine-readable description of all capabilities available for a WordPress site. This represented an expansion of infrastructure previously developed for the Abilities API and implemented with the release of WordPress 6.9 last December. A uniform approach to registration of capabilities by core components, by plugins, and by themes was implemented and extended for the May release of WordPress 7.0 (“Armstrong”) to include the WP AI Client.

Additional capabilities afforded by version 7.1 included custom hooks for validation of input and output, filtering of calls to wp_get_abilities, as well as a new action that executes for each call to an ability. The result was a system suitable for purposes of auditing, telemetry, tracing, and accounting for calls to abilities.

All calls to abilities could then be recorded by a site and represented the type of capability expected for future activity of agents. The major improvement represented a single flag for public access to capabilities. Activation of this flag for a particular ability provided immediate access to that ability via REST interfaces, via MCP adapters, and via the WP AI Client equivalent to implementation of a single flag for access to three types of service interface.

Previous exposure of a plugin to an external agent required manual implementation of a REST interface and hope that the developer of the agent had seen the documentation. With the present release, control of the conditions of this agreement was achieved by the platform.

The biggest change in 7.1 may be the one that most site owners never see: The new icons API allows registration of individual sets of SVG icons for server-side rendering and provides access to these sets through the architecture of the WordPress REST interface. This is yet another example of turning previously private capabilities into public, queryable features of the platform.

The release also fulfilled the long-standing objective of making "AI-ready" a major capability of the WordPress platform. Plugin developers, agencies, and other WordPress businesses should start adapting to these capabilities now. However, the platform also provides a way to control which capabilities are accessible to external agents.

The fully encapsulated editor is an absolute boundary, not a suggestion. All blocks that continue to be developed with the Block API at v2 or earlier must be converted to v3 to maintain compatibility as described by Kinsta.

Those that attempt to access components of the administrative interface as if they were on the same page will immediately lose all functionality upon activation of the fully encapsulated editor. WordPress VIP made its warning in this regard explicit in its availability statement to enterprise customers: "If your custom blocks or plugins rely on the editor document boundary, review that code before 7.1 makes it to your production sites." 7.1 is a release for the staging cycle, not for an afternoon of rapid clicks.

Your plugins that expose agent-accessible functions or register custom icons will require modifications to support the new interfaces, and your older blocks must be validated for Version 3 prior to any changes in production. For users of managed hosting, the appropriate periods of time ultimately will be determined for you and for many other users as well.

Planning ahead for the time required for your upgrade will always be the most economical approach. This isn't an excuse for neglecting the release, but reflects the cost of changing the foundation of a platform on which all users depend.

Changes in the administrative bar and in the editor will be the most obvious to users and editors during the fall. The most significant effects on machines will come from the use of the Abilities API. And interactions with software that never attended a WordCamp may ultimately be what "Mary Lou" is remembered for.