Webiny CMS MCP. Manage your whole content lifecycle via chat.

You're running a headless CMS—a big deal—and you don't want to play with some clunky GUI just to update a few fields. This MCP Server connects your AI agent straight into your Webiny content infrastructure, treating the whole thing like a set of robust APIs exposed through natural conversation. You'll manage content models, drafts, and published data using simple commands or raw GraphQL queries; you don't have to touch a playground interface.

Content Lifecycle Management: Getting Stuff Live (or Deleting It)

The server lets your agent handle the full life cycle of any piece of content. If you need to start writing something new, it runs create_cms_entry, which immediately makes an unpublished draft for whatever content model you specify. When that draft is ready, it uses update_cms_entry to modify specific fields on that existing revision—you don't have to rebuild the whole record just to fix a typo or change a date.

Once everything looks good, the agent promotes it using publish_cms_entry, making the content live and visible via the public Read API. If you ever need to clean house and permanently remove an old piece of content, delete_cms_entry takes care of that revision. Just remember: this action is irreversible.

Discovery and Auditing: Knowing What You're Dealing With

You can’t manage what you don’t know. For discovery, the agent first calls get_api_introspection. This fetches the entire GraphQL schema definition for your Webiny instance, giving your AI client a complete map of every data structure, type, and field available in your CMS environment. Need to see what models exist? You can run list_model_entries, which lists all existing content entries—you can tell it if you want live versions or just the drafts sitting in the system.

When you pinpoint a specific piece of content by its revision ID, get_model_entry_details pulls every single bit of metadata and field value associated with that entry, letting you audit exactly what's stored there. For high-level checks on the entire environment, running get_tenant_config pulls all the global configuration settings governing your whole Webiny tenant.

Raw Data Execution: When Simple Commands Won't Cut It

Sometimes, what you need is more power than a simple 'update' command offers. That’s where raw GraphQL comes in. The server exposes execute_graphql_query, allowing your agent to run any custom, complex mutation or query against the CMS API. You specify whether you're reading data, managing it, or just previewing it, and you can even set specific locales for those queries.

This tool means there’s no operation too complicated for your AI client—it executes whatever GraphQL structure you define.

It all works together. Your agent uses introspection to find the model (e.g., 'Products'). It lists existing entries to check the current count. Then, if it needs to modify a specific product's description and price, it runs update_cms_entry. If that change is ready for the public, it hits publish_cms_entry. If you need to pull every field from five different models into one spreadsheet, you just write a complex query using execute_graphql_query.

This whole setup gives your agent full, deep access to everything running in your CMS.