@actuate-media/mcp-server

A Model Context Protocol server that lets AI agents (Claude, Cursor, ChatGPT, etc.) read and write content in your Actuate CMS.

Install & connect

You don't need to install this globally — your MCP client will run it via npx.

Cursor / Claude Desktop:

{
  "mcpServers": {
    "actuate": {
      "command": "npx",
      "args": ["-y", "@actuate-media/mcp-server"],
      "env": {
        "ACTUATE_BASE_URL": "https://your-site.com",
        "ACTUATE_API_KEY": "act_sk_..."
      }
    }
  }
}

Generate an ACTUATE_API_KEY from your CMS admin under API Keys → New API Key. Grant it the scopes the agent needs (e.g. content + media + pageBuilder).

Tools exposed

| Tool | What it does | | ------------------------------ | ---------------------------------------------------- | | list_collections | Discover available collections and their fields | | list_documents | Paginated list of documents in a collection | | get_document | Fetch a document by ID | | create_document | Create a new document with explicit fields (DRAFT) | | update_document | Patch an existing document | | delete_document | Soft-delete a document | | get_global / update_global | Read/update globals (site settings, etc.) | | list_media | Browse the media library | | generate_page | AI page-builder: prompt → page tree (no persistence) | | create_page_from_prompt | One-shot: prompt → tree → persisted page document | | generate_block | AI generate a single page-builder block | | resolve_url | Map a public URL path to its CMS document |

AI content authoring (any non-page collection)

The agent can also author structured content for blog posts, case studies, services, locations, team profiles, FAQs, testimonials — any collection that isn't a page builder. Each tool maps a free-form prompt onto the collection's field schema and persists the result.

| Tool | Default collection | Notes | | ---------------------- | ------------------ | --------------------------------------------------------------------------------------------------- | | create_in_collection | (required arg) | Generic. Pass template: service \| location \| team_member for built-in prompt-enrichment recipes | | create_blog_post | posts | Blog post sugar | | create_case_study | case-studies | Plus clientName / industry hints |

All authoring tools share the same base inputs (prompt, tone, targetAudience, context, title, slug, publish) and accept a collection override when your site uses non-conventional slugs.

The standalone create_service / create_location / create_team_member tools were collapsed into create_in_collection. Use create_in_collection({ template: "service" | "location" | "team_member", ... }) instead.

Building pages section-by-section

Pages are authored as an ordered list of sections (the same model the visual Page Editor reads/writes), so an agent can build a real, previewable, editable page rather than disposable HTML.

| Tool | What it does | | ----------------------------------------------- | ------------------------------------------------------- | | list_section_types / get_section_schema | Discover registered section types + their field schema | | validate_section_content | Validate content against a type schema (no persistence) | | list_page_sections | List a page's ordered sections | | add_page_section / update_page_section | Add or patch a single section in place | | reorder_page_sections / remove_page_section | Reorder / remove a section | | toggle_page_section_visibility | Show or hide a section without deleting it |

Resolving relationships

Relationship fields store a target document ID, but prompts speak in slugs ("set the author to jane-doe"). These tools let an agent resolve them instead of guessing IDs.

| Tool | What it does | | --------------------------- | ----------------------------------------------------------------------------------- | | get_relationship_fields | List a collection's relationship/media/select fields + each field's relationTo | | resolve_relationship | Resolve an exact slug → { id, slug, title, status } within a collection | | list_relationship_targets | Browse candidate documents ({ id, slug, title, status }) when the slug is unknown |

Typical flow: get_relationship_fields(collection) → find the field's relationTo → resolve_relationship(relationTo, slug) → update_document with the returned ID.

Navigation menus

Menus live in the navigations collection (link tree on data.items). These tools wrap the document API with tree-aware conveniences so an agent builds a real, editable menu — and can wire a published page into one without hand-deriving its URL.

| Tool | What it does | | --------------------------------------------------------------- | ---------------------------------------------------------------------------------------------------------------- | | list_navigations / get_navigation | List menus ({ id, name, slug, location, itemCount }) or read one + its tree | | create_navigation / update_navigation / delete_navigation | Create, edit metadata / replace tree, or soft-delete a menu | | add_navigation_item | Add a link — explicit { label, url } or a page ref ("link page X into menu Y"); optional parentId nests it | | remove_navigation_item / reorder_navigation_items | Remove one item (+ sub-tree), or reorder one level by id permutation | | validate_navigation_tree | Validate a menu or candidate items (no save): { valid, errors, warnings, stats } |

Forms

The agent can author forms and manage the submissions inbox end-to-end (write-role API key required).

| Tool | What it does | | ---------------------------------------------------------------------------- | ------------------------------------------------------------------------- | | list_forms / get_form | Discover form definitions and read one | | create_form / update_form / delete_form | Create, edit metadata, or delete a form | | duplicate_form | Copy a form into a new draft | | get_form_schema / save_form_schema | Read or replace the (versioned) field schema | | list_form_submissions / list_form_entries / get_form_entry | Read submissions (PII-sensitive; explicit formId required) | | set_form_entry_status / bulk_set_form_entry_status / delete_form_entry | Manage inbox state (read/star/archive) and delete entries | | export_form_entries | Export all submissions as CSV (bulk PII egress; requires confirm: true) |

Scoping

The agent inherits whatever scopes you grant to its API key. Use the Content preset to allow read/create/update across all collections without admin powers. Use Custom to lock down to specific collections.

Direct (non-MCP) use

You can also import the underlying ActuateClient programmatically:

import { ActuateClient } from '@actuate-media/mcp-server'

const cms = new ActuateClient({
  baseUrl: 'https://your-site.com',
  apiKey: process.env.ACTUATE_API_KEY!,
})

const page = await cms.createPageFromPrompt({
  prompt: 'A landing page for a New York City house cleaning service',
  publish: false,
})