elementor-mcp
Server Configuration
Describes the environment variables required to run the server.
| Name | Required | Description | Default |
|---|---|---|---|
| WORDPRESS_URL | No | URL of the WordPress site (used for single-site configuration) | |
| WORDPRESS_USERNAME | No | Username for WordPress (used for single-site configuration) | |
| ELEMENTOR_MCP_SITES | No | JSON array or object defining multiple site profiles | |
| ELEMENTOR_MCP_CONFIG | No | Path to a JSON configuration file containing site profiles | |
| WORDPRESS_APP_PASSWORD | No | Application password for WordPress (used for single-site configuration) | |
| ELEMENTOR_MCP_READ_ONLY | No | If set to 'true', refuses all write operations across all sites | false |
| ELEMENTOR_MCP_TIMEOUT_MS | No | Per-request timeout in milliseconds | 30000 |
| ELEMENTOR_MCP_DEFAULT_SITE | No | Name of the default site to use when multiple sites are configured |
Instructions
Guidance the server publishes about itself, which clients place ahead of the tool catalog so the model reads it before choosing anything.
Tools for reading and editing Elementor pages on WordPress sites.
Working order that avoids the usual mistakes:
elementor_site_status once per site, to see what Elementor version and widgets are available.
elementor_get_outline to get element ids — do not fetch whole element trees unless you need them.
elementor_get_widget_schema before setting unfamiliar widget settings; control names are not guessable.
elementor_batch_edit for multi-step changes, so the page cannot be left half-edited.
Layout: check containerAvailable in elementor_site_status before building. Elementor gates the flexbox container behind an experiment that is off by default on sites installed before 3.16, so a current Elementor does not guarantee containers exist. Where it is false, build with section and column instead.
Every write snapshots the page first; elementor_restore_snapshot undoes the last change. Writes carry the hash the page was read at and are rejected if it changed underneath you. Site-wide replace defaults to a dry run — read the report before committing.
Capabilities
Features and capabilities supported by this server
Protocol revision2025-11-25
| Capability | Details |
|---|---|
| tools | {
"listChanged": true
} |
| prompts | {
"listChanged": true
} |
Tools
Functions exposed to the LLM to take actions
| Name | Description |
|---|---|
| elementor_list_sitesA | List the WordPress sites this server can reach, with which one is the default. Call this first when you are unsure which site name to pass to other tools. |
| elementor_site_statusA | Report what a site is running: WordPress and Elementor versions, whether Elementor Pro is present, how many widget types are registered, the active kit id, the capabilities of the authenticated user, and whether the bridge plugin and destructive operations are enabled. Run this before a first editing session on an unfamiliar site. |
| elementor_native_mcp_callA | Invoke one of Elementor's own built-in MCP abilities through the bridge. Elementor 4.3+ ships abilities covering v4 atomic features (global classes, variables, components, compositions) that have no v3 equivalent. Only works where that module is active — check elementor_site_status first, under nativeMcp. |
| elementor_list_pagesA | List WordPress pages, posts and templates, newest edited first, with whether each is built with Elementor. Use elementorOnly to skip content the page builder does not own. |
| elementor_get_outlineA | Read the structure of an Elementor page: every element id, its type, and a short label taken from its most identifying setting. This is the tool to reach for first when editing a page — it is a small fraction of the size of the full element tree and gives you the ids every other element tool needs. It also returns the content hash used for safe concurrent writes. |
| elementor_get_pageA | Read a page: metadata, a census of which widgets it uses, its content hash, and optionally the full element tree. Leave includeElements false unless you genuinely need every setting — element trees run to hundreds of kilobytes on real pages. Prefer elementor_get_outline, then elementor_get_element for the specific parts you care about. |
| elementor_create_pageA | Create a new Elementor-enabled page, post or template. Starts empty unless you pass an element tree. Created as a draft by default so nothing goes live unreviewed. For a full-width blank canvas set pageTemplate to elementor_canvas. |
| elementor_update_page_metaA | Change a page's title, slug, status, parent or featured image. This edits WordPress post fields, not the Elementor layout — use the element tools for that. |
| elementor_duplicate_pageA | Copy a page, its Elementor layout and its page settings into a new draft, regenerating every element id so the copy is independent. The usual way to iterate on a live page safely: duplicate it, edit the copy, then publish when it is right. |
| elementor_delete_pageA | Move a page to the trash, or delete it permanently with force. Permanent deletion additionally requires the site to have opted into destructive operations and confirm to be true. |
| elementor_render_pageA | Render a page — or one element of it — to HTML as the front end would. Use this to check what an edit actually produced, especially for widgets whose output depends on dynamic data. |
| elementor_get_page_treeA | Read a page's complete element tree including every setting. This is the largest thing this server returns — only call it when you need to inspect settings across many elements at once. For normal editing use elementor_get_outline plus elementor_get_element. |
| elementor_get_elementA | Read a single element and its subtree, including every setting. Use this after elementor_get_outline to inspect just the part of a page you are about to change. |
| elementor_add_widgetA | Insert a new widget into a page. Pass the widget type (heading, image, button, text-editor, and so on — see elementor_list_widgets) and its settings. With no targetId the widget is appended to the page root; a widget must otherwise go inside a container or column, so use position before/after to place it beside an existing widget. |
| elementor_add_containerA | Insert a structural element — a flexbox container (the modern default), or a section/column for pages built the classic way. Containers are what you put widgets inside. You can seed it with child elements in one call. |
| elementor_update_elementA | Change an element's settings. In merge mode (the default) the keys you pass are applied over the existing settings and everything else is left alone; in replace mode the settings object is swapped wholesale, which discards any setting you do not include. Merge is almost always what you want. |
| elementor_move_elementA | Relocate an element within the page. Give a referenceId plus a position to say where it lands; omit referenceId to move it to the page root. Moving an element into its own descendant is refused. |
| elementor_duplicate_elementA | Copy an element and its whole subtree, placed immediately after the original, with fresh ids throughout. The quickest way to repeat a card, column or row you have already styled. |
| elementor_delete_elementA | Remove an element and everything inside it from a page. A snapshot is taken first, so elementor_restore_snapshot can undo it. |
| elementor_reorder_childrenA | Set the order of a container's direct children by listing their ids. Ids you leave out keep their relative order at the end, so you can move one item to the front without listing everything. |
| elementor_wrap_elementA | Put a new container around an existing element, in place. Useful for adding a background, padding or width constraint around something without rebuilding it. |
| elementor_batch_editA | Run a list of element operations against a page as one atomic change: insert, update, move, duplicate, delete, reorder and wrap. Operations apply in order, each seeing the result of the last. If any one fails the whole batch is abandoned and nothing is written. Prefer this over a series of single-element calls when building or restructuring a section — it is one read and one write instead of many, and it cannot leave the page half-changed. |
| elementor_replace_page_treeA | Overwrite a page's whole element tree. This discards the existing layout completely, so reach for it when generating a page from scratch rather than to make a change. A snapshot is taken first. For edits to an existing page use elementor_batch_edit instead. |
| elementor_find_elementsA | Search one page for elements by widget type, by text in their settings, or both, and get back their ids with a short label. Use this to locate what you want to edit without reading the whole tree. |
| elementor_list_widgetsA | List every widget type registered on the site, with its title, panel categories and whether it comes from Elementor Pro. Search by name or keyword to find the right one. Always check here before adding a widget — a site only has the widgets its installed plugins provide. |
| elementor_get_widget_schemaA | Read the full control schema for a widget or structural element: every setting it accepts, grouped by panel tab and section, with types, defaults, allowed options and which settings are responsive. Read this before setting unfamiliar widget settings — guessing control names is the most common cause of an edit that saves without visible effect. |
| elementor_list_element_typesB | List the structural element types available (container, section, column) and the widget panel categories the site defines. |
| elementor_list_dynamic_tagsA | List the dynamic tags registered on the site — post title, ACF field, site logo and so on — which let a widget setting pull live data instead of holding a fixed value. Mostly an Elementor Pro feature. |
| elementor_get_globalsA | Read the site's Elementor kit: global colour and typography palettes with the ids widgets refer to them by, layout defaults such as container width and breakpoints, and the site custom CSS. Read this before styling anything — using an existing global token keeps a site consistent, and changing one restyles every element bound to it. |
| elementor_set_global_colorA | Change one global colour by id (primary, secondary, text, accent) or by its title, or add a new custom colour if it does not exist. Every element bound to that token updates at once, and Elementor's CSS is regenerated. This is the right way to restyle a site's palette. |
| elementor_update_globalsA | Merge settings into the Elementor kit — colour and typography palettes, container width, breakpoints, theme style defaults. This changes the whole site at once, so read elementor_get_globals first and change only the keys you mean to. A snapshot of the kit is taken before the write. |
| elementor_custom_cssA | Read the site-wide custom CSS held on the Elementor kit, or replace it. Passing css writes it; omitting css just reads. Note that Elementor only renders this on Pro. |
| elementor_get_global_classesA | Read Elementor v4 global classes — the reusable style classes used by the atomic/Editor One element model. Returns an empty list on sites that predate that feature. |
| elementor_flush_cssA | Clear Elementor's cached CSS so it regenerates on the next page load. The editing tools already do this after every write; call it manually when styles look stale after changes made outside this server, or after a theme or plugin update. |
| elementor_list_templatesB | List the Elementor templates saved on this site — sections, containers, pages, headers, footers and popups. Reusing a proven template is cheaper and more consistent than rebuilding the same layout. |
| elementor_get_templateA | Read a template's element tree. By default ids are regenerated so the tree can be pasted into a page without colliding with anything already there. |
| elementor_save_templateA | Save a layout to the template library, either from an element tree you pass or by copying from an existing page (optionally just one element of it). Save a section once, reuse it everywhere. |
| elementor_apply_templateA | Insert a saved template's layout into a page, with fresh element ids. Place it at the page root, or relative to an existing element with targetId and position. The page is snapshotted first. |
| elementor_export_templateA | Export a template in Elementor's own JSON format, ready to import into another site with elementor_import_template or through the Elementor UI. |
| elementor_import_templateA | Import an Elementor template JSON payload as a new template on this site. Pass the decoded JSON object from an Elementor export or from elementor_export_template — which makes copying a layout between sites a two-call operation. |
| elementor_list_snapshotsA | List the layout snapshots stored for a page, newest first, with when each was taken and why. One is taken automatically before every write this server makes. |
| elementor_create_snapshotA | Take a named snapshot of a page's current layout. Worth doing before a large restructuring so you have one clearly labelled point to come back to. |
| elementor_restore_snapshotA | Roll a page back to a snapshot. With no snapshotId this restores the most recent one, which undoes the last change made through this server. The current state is snapshotted first, so a restore is itself reversible. |
| elementor_list_revisionsA | List a page's WordPress revisions, flagging which of them actually carry Elementor layout data. Revisions reach further back than snapshots but are less reliable for Elementor content. |
| elementor_restore_revisionA | Restore a page's Elementor layout from a WordPress revision. Only works on revisions flagged as carrying Elementor data. A snapshot is taken first. |
| elementor_searchA | Search inside Elementor page data across the site — by text, by widget type, or by which elements define a given setting. Elementor stores its content as JSON in postmeta, so it is invisible to normal WordPress search; this is the way to find "which page has that old phone number on it". |
| elementor_replace_textA | Replace a string everywhere it appears in Elementor page data — a phone number, an old brand name, a changed URL. Runs as a dry run by default and reports exactly which pages and how many occurrences it would touch; set dryRun false and confirm true to commit. Every page it edits is snapshotted first. Always read the dry run before committing. |
| elementor_widget_usageB | Count which Elementor widgets are actually used across the site and how much each page contains. Useful before removing an addon plugin, or to find the heaviest pages. |
| elementor_list_mediaB | Search the WordPress media library and get back attachment ids and URLs. Elementor image controls need both, in the shape {"id":123,"url":"https://..."}. |
| elementor_upload_imageA | Download a remote image into the WordPress media library and return the attachment id and URL, plus a ready-made value for an Elementor image control. Use this before setting any image setting — Elementor image controls want a library attachment, not a remote URL. |
Prompts
Interactive templates invoked by user choice
| Name | Description |
|---|---|
| edit_page_safely | The read-inspect-change-verify loop for changing an existing page. |
| build_page | Compose a new page from containers and widgets without breaking a live site. |
| audit_site | Survey what a site is built from before changing or migrating it. |
Resources
Contextual data attached and managed by the client
| Name | Description |
|---|---|
No resources | |
TDQS
Scored across 50 tools
The read tools are explicitly layered — get_outline for structure, get_element for one subtree, get_page_tree for everything, get_page for metadata — and the descriptions state when to use which, which is unusually good for a set this size. A few clusters still require careful reading (batch_edit vs replace_page_tree vs duplicate_page, and snapshot vs revision restore), but no two tools are truly interchangeable.
Everything is snake_case with a uniform elementor_ prefix, and the vast majority is verb_noun (list_sites, create_page, update_element, restore_snapshot). Minor deviations are noun-only names like elementor_custom_css and elementor_widget_usage, plus the opaque elementor_native_mcp_call, but the pattern stays predictable.
Fifty tools is a heavy surface even for a broad site-building domain, and several groups are redundant in scope: four ways to read page content, five snapshot/revision tools, and both duplicate_page and duplicate_element. A capable agent can navigate it, but the sheer count raises selection cost and invites over-fetching of large trees.
Coverage is strong end to end: page and element CRUD, structural wrapping/reordering, atomic batch editing, templates, snapshots and revisions, globals, media, and site-wide search/replace. The notable gaps are lifecycle gaps on templates (no update or delete, only save/export/import) and no media deletion or update, which agents will hit when managing the template library.