AI Contributor Weekly Summary – 4 February 2026


This week’s AI contributor meeting focused heavily on final preparation for the WordPress AI Client merge proposal, including provider separation, core integration strategy, and controls for disabling AI functionality. The group also reviewed progress across Abilities, MCP, and the AI Experiments plugin, with an eye toward unblocking work ahead of the WordPress 7.0 Beta 1 milestone on February 19.

WP AI Client Merge Proposal

@jason_the_adams shared a detailed update on the WordPress AI Client merge proposal and next steps toward core inclusion.

  • The WordPress AI Client will be fully provider- and model-agnostic when proposed for core.
  • Embedded providers are being removed from the PHP AI Client, with OpenAI, Google, and Anthropic providers moving into separate plugin repositories on WordPress.org.
  • @jason_the_adams plans to submit the core work as multiple PRs rather than a single all-or-nothing proposal, starting with the PHP AI Client, WordPress AI Client, and prompt builder.
  • Integration of the PHP AI Client into core remains the largest open question due to Composer dependencies and build considerations. The group discussed approaches similar to Gutenberg’s sync/build process, and @jason_the_adams plans to follow up with @youknowriad for guidance.

There was consensus that publishing a technical PR soon is critical to allow meaningful core review before Beta 1.

Disabling and Controlling AI Functionality

The group discussed approaches for disabling or restricting AI usage in WordPress.

  • @jason_the_adams introduced an initial filter-based approach that allows site owners to disable AI entirely or restrict usage to certain capabilities.
  • @justlevine raised concerns that filters alone are insufficient, as they can be overridden by plugins, and suggested introducing a constant to represent an explicit site-level decision.
  • @jason_the_adams agreed to add a constant-based control mechanism, while the group aligned that adding UI-level “kill switches” would be premature at this stage.

The goal is to make it easy for sites to opt out without introducing unnecessary UI or policy complexity.

AI Experiments Plugin

@jeffpaul shared updates on the AI Experiments plugin and near-term priorities.

  • Work is wrapping up on Featured Image Generation, Content Summarization, and Alt Text Generation in preparation for a 0.3.0 release.
  • Data Views and Data Forms UI work is being deferred until after the 7.0 Beta 1 window to keep focus on the AI Client merge proposal.
  • A Call for Testing post covering several experimental features has been published, and the team will track feedback before deciding whether any of those concepts should be milestone candidates.
  • Markdown-related work will remain in the experiments plugin for now to gather feedback and usage data before considering core or Gutenberg integration.

Abilities Update

@jorgefilipecosta provided a status update on Abilities work:

  • A PR allowing deeper nesting of abilities namespaces has been approved.
  • Updates to settings abilities and a new user ability (without meta support) are nearing readiness.
  • A revised post-finding ability that works across post types is in progress and may require additional discussion.
  • The group expressed optimism that settings and user abilities could land in WordPress 7.0.

This work also helps reduce immediate pressure around formal versioning by allowing namespacing patterns to evolve organically.WordPress well-positioned for future developments such as embedded models or host-managed AI offerings.

MCP (Model Context Protocol)

@jason_the_adams requested review on the final PR in a six-part MCP adapter update.

  • The work introduces a generated PHP schema layer derived directly from the MCP JSON schema.
  • The update decouples component registration from abilities, allowing MCP components that do not need to exist as abilities.
  • @jorgefilipecosta offered to review the remaining PR to help move the adapter forward.

Provider Discovery, Documentation, and Guidelines

Several forward-looking topics were discussed:

  • Provider discovery and validation: The group discussed how AI provider plugins should be discoverable without overwhelming users, including possible use of tags and compatibility checks rather than curated lists.
  • Documentation: Two documentation tracks were identified as necessary—one for plugin developers using the AI Client and prompt builder, and another for provider and host integrations.
  • Guidelines and disclosure: Contributors were encouraged to review @isotropic’s draft disclosure statement for agent skills as early prior art for applying the newly published AI Guidelines.

Next steps

  • @jason_the_adams will submit core PRs for the PHP AI Client and WordPress AI Client and add a constant-based mechanism for disabling AI.
  • @jeffpaul will continue coordinating the AI Experiments 0.3.0 release and track feedback from the testing call.
  • @jorgefilipecosta will review the remaining MCP adapter PR and continue advancing Abilities work.
  • Contributors are encouraged to review the MCP PRs and provide feedback on the agent skills disclosure draft.
  • Follow-up discussions on documentation, WP Bench, and provider discovery will continue in Slack.

Upcoming meetings

Folks are welcome to join on Wednesday’s at 1700 UTC via Google Meet with in-meeting notes captured in a Slack Canvas and then paired with Gemini meeting notes to help generate this meeting summary post. All team meetings are published to https://make.wordpress.org/meetings/#ai.

  • The next bi-weekly AI Team Chat Slack discussion is scheduled for 5 February 2026.
  • The next weekly AI Contributor weekly Google Meet video call is scheduled for 11 February 2026.

Props to @psykro for pre-publish review.

#core-ai, #summary