SectionMate MCP Wordpress Plugin - Rating, Reviews, Demo & Download
Plugin Description
Page designer, not website designer. This plugin builds and edits page content through page builders (Elementor, Gutenberg, Spectra). It does not configure WordPress themes, theme builders, or the global site header/footer. You can design a nav or hero “header” on the page as blocks; that is not the same as installing a theme site header (which often needs a theme’s pro site builder — intentionally out of scope so we stay portable across themes).
Do not leave the MCP connector unattended on a live site. A connector acts as your account and can do what your account can do — no more — and any one of them can be revoked from the Setup screen without affecting the others or your login. The bigger risk is judgement rather than access: an AI editing a live page is still an AI editing a live page. Prefer local or staging for large redesigns. On production, prefer draft pages and short sessions; turn the MCP connector switch off under Setup when finished (no need to uninstall) — compiled page content stays.
SectionMate MCP lets AI agents design and edit pages over an MCP connector. Agents work on compact clean JSON; the plugin compiles that JSON into valid Spectra (Legacy + V3), Gutenberg, and Elementor page output — so each turn uses fewer tokens and fewer retries.
What you get
- Design with clean semantic JSON; compile to Spectra blocks, core Gutenberg blocks, or Elementor trees
- On-page sections including nav/hero “header bands” as page content (not theme site headers)
- Optional hide theme header/footer on a landing page (
hideHeaderFooter) — does not replace the theme header - Decompile existing pages back to the same clean JSON shape
- Live MCP tools: list pages/blocks, preview edits, then confirm before anything is saved
- Front-end Inspect and Annotations for design review with a logged-in editor
- Built-in skill guide the AI can load with
get_skill - After you deactivate the plugin, page content remains (blocks, Elementor data, and any CSS/JS already saved into the page)
What this is not
- Not a theme installer or full-site / FSE site designer
- Not Astra Site Builder, Header Footer Builder, or any theme’s global header system
- Not a substitute for Customizer menus and theme chrome — focus stays on page builders and page content
How the connector works
- Activate the plugin (local, staging, or a careful production session)
- Open SectionMate MCP in the admin sidebar, type a name for the AI client you are connecting (“Claude Desktop”, “Grok”), and click Create connector URL
- Copy the URL it shows you. It already contains your username and an Application Password — there is nothing to assemble. It is shown only once, because WordPress stores the password encrypted and it cannot be displayed again. Lose it and you simply revoke that connector and create another.
- Expose local sites with Cloudflare Tunnel or ngrok if needed (Application Passwords require HTTPS)
- Paste the URL into your AI client as a remote MCP / connector server. Clients that let you set headers can send
Authorization: Basicwith the plain URL instead. - Ask the AI to design or edit pages — every write is preview, then confirm
- When finished: turn the MCP connector switch off under Setup (recommended), or revoke that connector
The Setup screen lists every connector with access, when it was created, and when it was last used — which is how you spot one you forgot to turn off. Revoking one does not affect the others or your own login.
The connector acts as you. It can do what your account can do and no more, and every edit is attributed to your user.
Tips for better AI designs
- Section by section — finish and review one band (hero, features, FAQ…) before the next. Full-page builds work, but more blocks and decisions use more context and can weaken layout quality.
- Feed real direction — attach screenshots, brand colours, logos, backgrounds, and use media-library images so the AI has concrete references.
- Ask for clean reading — request clear hierarchy, short paragraphs, and scannable structure; the AI can produce strong typesetting for long-form text when you prioritize readability.
Treat the connector URL like a password — it contains one. Anyone who has it can act as your account on that site. Revoke it at any time from the plugin’s Setup screen (or from Users Profile Application Passwords); doing so does not affect your normal login or the other connectors.
Important safety notes
- Session, not always-on. Use the admin MCP connector on/off switch when you finish. Do not leave production with the connector on 24/7. Prefer staging for large redesigns; on live sites use drafts and short sessions.
- The MCP URL contains an Application Password. It grants exactly your account’s permissions — no more — and is revocable on its own.
- The plugin writes no remote script tags. Nothing it compiles loads code from another server, and there is no setting that changes that. Animations are plain CSS and DOM APIs written into the page.
- Custom JavaScript follows WordPress’s own rule. A
codeblock can carry your CSS and JavaScript, but writing a live<script>into page content requires theunfiltered_htmlcapability — the same capability core requires for the Custom HTML block. On a single site WordPress grants it to administrators and editors; on multisite only super admins have it. Connect as an account without it andcodenodes keep their CSS and HTML while the JavaScript is filtered out, exactly as if you had pasted it into the editor by hand. - Inline SVG icons survive that filtering: the plugin extends the allowed-tags list to permit SVG shape elements (no
script, no event handlers, nohref), so icons still render. One caveat for accounts withoutunfiltered_html: WordPress lowercases attribute names while filtering, soviewBoxis stored asviewbox. Browsers correct this when rendering, but the block editor may flag such a block as containing unexpected content.
Compatibility
Works with:
- WordPress block editor (Gutenberg)
- Spectra / Ultimate Addons for Gutenberg (UAGB), when installed
- Spectra Blocks (V3), when installed
- Elementor free, when installed
Not affiliated with or endorsed by Elementor, Brainstorm Force (Spectra), or Automattic.
Supported blocks & elements
The AI writes one clean JSON shape. The plugin compiles it into the blocks or widgets you already have available. Most good pages only need the foundation five — and those are the deepest mappings on every backend:
- Containers / sections / grids — layout bands, nested rows, multi-column grids
- Headings & text — titles, hierarchy, body copy
- Images — hero and content images from the Media Library or imported URLs
- Buttons — primary and secondary CTAs
- Icons — feature rows and list accents
Spectra Legacy (UAGB) when installed: uagb/container, uagb/advanced-heading, uagb/image, uagb/icon, uagb/buttons (+ child), plus counter, countdown, FAQ, icon-list, image-gallery. Other uagb/* blocks can be edited as attribute shells. Optional code nodes use core/html for scoped CSS/JS.
Spectra Blocks V3 when installed: spectra/container, spectra/content (heading + paragraph), spectra/buttons + spectra/button, spectra/icons + spectra/icon, core/image, spectra/counter, spectra/countdown, spectra/accordion (FAQ), spectra/list.
Gutenberg (core only): core/group, core/columns + core/column, core/heading, core/paragraph, core/buttons + core/button, core/image, core/gallery, core/list, core/html. FAQ and counter compile as static compositions (not live animated widgets).
Elementor free when installed: Container, Heading, Image, Icon, Button, HTML, Counter, Countdown, Icon List, Accordion (FAQ), Basic Gallery / Carousel. Unmapped widgets decompile as shells and are preserved, not re-authored.
Do not mix backends on the same page. Agents can call list_block_types, get_block_schema, list_widget_types, and get_element_schema for live field maps.
Privacy
The plugin does not phone home, and it collects no analytics or telemetry of any kind. It adds no third-party scripts and no trackers to your pages.
It does use two external services, and one of them (Google Fonts) is requested by your visitors’ browsers. Both are described in External services below. Read that section before using the plugin on a site with GDPR or similar obligations.
External services
This plugin relies on the following external services. Each is listed with what it does, what is sent, when, and links to that provider’s terms and privacy policy.
Google Fonts (fonts.googleapis.com, fonts.gstatic.com)
What it is used for. When a page you compiled specifies a Google font family, the plugin enqueues that family’s stylesheet so the page renders in the font you chose. The plugin’s own admin screen and the logged-in front-end Inspect overlay also load one font (IBM Plex Mono) for their interface.
What is sent and when. The plugin builds a stylesheet URL containing only the font family names and weights the page asks for. Your visitor’s browser then requests that stylesheet from fonts.googleapis.com and the font files from fonts.gstatic.com. Those requests are made by the browser, so Google receives the visitor’s IP address and user-agent, as it does for any site using Google Fonts. This happens on the front end for pages that use a Google font, and in the admin/editor for the interface font. No personal data is sent by the plugin itself, and nothing is sent when a page uses no Google font.
Avoiding it. Pages that use only system or theme fonts make no request. If you need zero third-party font requests, do not specify Google font families when designing, and use a font-hosting plugin for the interface font.
Service provided by Google: terms of service, privacy policy, Google Fonts privacy FAQ.
Wikimedia Commons API (commons.wikimedia.org)
What it is used for. The optional search_placeholder_images MCP tool searches Wikimedia Commons for openly licensed photographs, so an AI agent can suggest real images for a design instead of inventing image paths that render blank. It is read-only and returns image URLs, dimensions, suggested alt text, and licence/attribution information.
What is sent and when. Only when you or your connected AI agent explicitly calls that tool. The request is made from your server and contains the search terms supplied in the call, plus a requested image width and result limit. No personal data, no site information, and no account or API key are involved.
Avoiding it. The tool only runs on request. If it is never called, the plugin never contacts Wikimedia. Turning the MCP connector off under Setup blocks it along with every other tool.
Service provided by the Wikimedia Foundation: terms of use, privacy policy, API documentation.
Images you choose to import (no fixed provider)
What it is used for. The import_media tool, and the media sideloading step of compile_and_insert / compile_elementor, download an image from a URL and add it to your Media Library so the page serves it locally rather than hotlinking.
What is sent and when. Your server makes an HTTP request to whatever URL you or your agent supplies — there is no fixed third party, and the plugin has no default host. It runs only on explicit request, requires the upload_files capability, and is confirm-gated: the first call returns a preview and nothing is downloaded until a second call confirms it. Whichever host the URL points at will see a request from your server’s IP address.
Avoiding it. Supply only URLs you trust, or upload images through the WordPress Media Library instead and reference them by attachment id.
Screenshots
Admin screen with safety warning and MCP connector URL
Setup steps for connecting an AI client
Front-end Inspect / design review for logged-in editors
