WordPress MCP Write Capabilities: AI Agents Can Now Write, Edit, and Publish on 43% of the Web

WordPress powers 43% of all websites on the internet. On March 20, 2026, Automattic flipped a switch that lets AI agents — Claude, ChatGPT, Cursor, and any MCP-compatible client — not just read WordPress.com sites, but write to them. Draft blog posts. Build landing pages. Manage comments. Restructure categories. Fix SEO metadata. All through natural language conversation.

This isn’t a plugin or a third-party hack. It’s a first-party feature built into WordPress.com’s infrastructure using the Model Context Protocol (MCP), the open standard Anthropic introduced to let AI systems interact with external tools. And given WordPress’s 60.5% CMS market share, the ripple effects could be enormous.

What the 19 Write Operations Actually Do

Back in October 2025, WordPress.com shipped read-only MCP support — AI agents could pull site data but couldn’t change anything. The March 2026 update adds 19 write operations across six content types, turning passive reading into active content management.

Posts and Pages are where the headline feature lives. An AI agent can draft a blog post, build an About page with team bios and contact info, or create a landing page that automatically inherits the site’s theme — colors, typography, spacing, and block patterns included. The agent reads the design system before generating content, so outputs aren’t generic templates but theme-aware compositions.

Comments get full moderation capability. An agent can approve pending comments, reply to them, or delete spam — all from a single natural language prompt like “approve all comments from the last week and reply to the ones asking questions.”

Categories and Tags support creation, renaming, and hierarchical restructuring. You can tell an agent to “set up a Recipes category with subcategories for Breakfast, Lunch, Dinner, and Desserts” and it builds the taxonomy tree. It also checks for existing categories to avoid duplicates.

Media Management covers metadata, not file uploads. Agents can audit image libraries for missing alt text, fix captions, and update titles — practical SEO housekeeping that most site owners never get around to doing manually.

The technical interface uses an action-based design: “list” to discover available operations, “describe” to get the schema for a specific one, and “execute” to run it. This makes MCP write capabilities discoverable by any compatible AI client without hardcoded integrations.

The Safety Architecture: Drafts, Approvals, and Activity Logs

The most common reaction to “AI can now publish to WordPress” is concern — and Automattic clearly anticipated it. The safety model has multiple layers.

Every single write operation requires explicit human approval. Before creating, updating, or deleting anything, the agent describes exactly what it plans to do and asks for confirmation. No silent background publishing. No surprise edits to live pages.

New posts default to draft status. Even after you approve an agent’s action, the content lands as a draft, not a published page. If an agent modifies an already-published post, it triggers a warning that changes will be immediately visible.

Deletions are recoverable — mostly. Posts, pages, comments, and media go to trash with a 30-day recovery window. Categories and tags are the exception: WordPress doesn’t support trashing taxonomies, so deleting them is permanent. The agent is required to issue an additional warning and get extra confirmation before proceeding.

WordPress role permissions apply. If your account is a Contributor, the agent can draft but not publish. If you’re an Editor, the agent can modify posts but not site settings. The AI doesn’t get elevated privileges — it operates within the same permission boundaries as the human user.

Everything logs to the Activity Log. Every agent action is recorded, so site owners can audit what happened and when. Users can also ask their agent for a summary of recent changes at any time.

Granular toggles let site owners enable or disable specific operations independently. Don’t want agents touching comments? Turn off comment write access while keeping post creation enabled.

How WordPress MCP Write Capabilities Stack Up Against the Competition

WordPress isn’t the only CMS betting on AI, but its approach is fundamentally different from what Wix, Squarespace, and others are doing.

Wix has invested heavily in AI website generation — its chatbot asks questions about your business, style preferences, and goals, then produces a personalized site. It’s a polished onboarding experience, but the AI is Wix’s AI, running inside Wix’s ecosystem. You don’t bring your own agent.

Squarespace takes a similar path with Blueprint AI, offering guided AI tools for design, SEO, and copywriting. Its Brand Identity feature lets you describe your business personality so AI-generated text stays on-brand. Again, it’s a closed system — Squarespace’s tools, Squarespace’s models.

WordPress MCP Write Capabilities take the opposite approach: they don’t ship their own AI model. Instead, they expose a standardized protocol that any MCP-compatible agent can connect to. Want to use Claude for drafting and Cursor for code-heavy landing pages? You can. Want to switch from ChatGPT to a local model next month? The MCP interface stays the same.

This open-protocol strategy aligns with WordPress’s historical advantage — extensibility. But it also creates a unique risk profile. When the AI isn’t your own, quality control depends entirely on the user’s choice of agent and the prompts they write. The platform provides the pipes; what flows through them is up to the user.

Market dynamics add context here. Between December 2024 and December 2025, WordPress contracted 2.9% in market share while Wix grew 22.4% and Squarespace grew 6.2%. Betting on MCP as an open standard, rather than building a proprietary AI experience, could be WordPress’s strategy to differentiate in a market where closed AI integrations are becoming table stakes.

The Bigger Question: What Happens When AI Can Publish at Scale?

The practical upside is obvious. A solo entrepreneur running a WordPress site can now tell Claude to “draft a 1,000-word post about our new product launch, categorize it under Announcements, add relevant tags, and write a meta description under 160 characters.” The agent handles the grunt work. The human reviews and hits publish. Time saved, output increased.

But TechCrunch’s coverage flagged the tension directly: these AI agents “could lower barriers to publishing while increasing machine-generated content across the web.” When the platform that powers nearly half the internet makes it trivial for AI to publish content, the volume implications are staggering.

Google has spent the last two years refining algorithms to identify and demote AI-produced pages that lack genuine value. Content creators and journalists have raised concerns about an already-polluted web getting worse. And there are legal gray areas — who’s accountable when an AI agent publishes defamatory or factually incorrect content? The technology is moving faster than regulatory frameworks can follow.

The WordPress community itself is split. Some see MCP write capabilities as a way to do more with less — small teams punching above their weight with AI assistance. Others worry about quality control at scale, and some developers have expressed anxiety about investing in MCP infrastructure while the technology and implementation are still maturing.

Automattic’s bet seems to be that the safety architecture — mandatory approvals, draft defaults, role-based permissions, activity logging — is sufficient to keep humans in the loop. Whether that holds up when thousands of sites start running AI content pipelines is an open question.

Setting It Up

The setup is straightforward for existing WordPress.com users:

Navigate to wordpress.com/me/mcp
Toggle on the write capabilities you want (posts, comments, categories, tags, media)
Connect your preferred MCP client — Claude, ChatGPT, Cursor, or any compatible tool
Start managing your site through conversation

Write capabilities are available on all WordPress.com paid plans at no additional cost. No plugins to install, no server configuration required. For Cursor users, there’s a one-click integration through the Cursor MCP Directory.

Worth noting: this is WordPress.com (the hosted platform), not self-hosted WordPress.org. However, the WordPress MCP Adapter is moving toward WordPress Core inclusion, which could eventually bring these capabilities to the broader self-hosted ecosystem.

FAQ

Is WordPress MCP Write Capabilities free?
Write capabilities are included on all WordPress.com paid plans at no extra charge. You do need a paid plan — the free tier doesn’t include MCP access. The cost of the AI agent itself (Claude, ChatGPT, etc.) is separate and depends on your subscription with that provider.

Which AI agents work with WordPress MCP Write Capabilities?
Any MCP-compatible client works. The officially confirmed clients include Claude (Anthropic), ChatGPT (OpenAI), and Cursor. Since MCP is an open protocol, other tools that implement MCP support should also be compatible.

Can AI agents publish content without my approval?
No. Every write operation requires explicit human confirmation before execution. New posts default to draft status even after approval. Modifications to published content trigger additional warnings. The AI agent cannot bypass these approval requirements.

How does WordPress MCP Write Capabilities compare to Wix and Squarespace AI features?
The key difference is architectural. Wix and Squarespace offer proprietary AI tools built into their platforms — you use their AI, their way. WordPress MCP exposes a standard protocol that any compatible AI agent can connect to, giving users freedom to choose their preferred AI tool. The tradeoff is that WordPress doesn’t control the AI quality; that depends on which agent you use.

Will this work on self-hosted WordPress.org sites?
Not yet. The current MCP write capabilities are exclusive to WordPress.com hosted sites. However, the WordPress MCP Adapter is being developed with a path toward WordPress Core inclusion, which would eventually extend these capabilities to self-hosted installations. Third-party MCP server implementations for self-hosted WordPress also exist on GitHub, though they aren’t officially supported by Automattic.

Top AI Product

Leave a comment Cancel reply