Djordje Savanovic, view all articles
10.11.2026
14 min read

Claude Integration in Strapi How we connected our CMS with Claude

  • AI
  • Artificial Intelligence
  • cms
  • content
  • neue Technologie
Short summary:

Content teams spend a large part of their time on writing tasks: creating entries, filling fields, maintaining metadata, and copying texts between tools. This is exactly where our Claude Integration comes in. At LEAN-CODERS, we connected our Strapi CMS directly to Claude via an MCP server. Claude works on the content where it resides, and the team retains control.

In this article, we will show how the connection is established, what role Claude plays in the editorial daily routine, and what costs and benefits you can expect.

What you'll find in this article:

Why we now use a Claude integration over MCP for our Strapi CMS

A CMS is the place where content is created, reviewed, and published to a website or other applications. Those who want to use AI for this quickly face a media break: The text is created in a chat window, copied, pasted into the CMS (in our case Strapi), formatted, and provided with metadata. Each of these steps takes time and is a potential source of error.

We wanted to eliminate this break and perhaps also offload some time-consuming formatting tasks. The goal was for Claude to work directly with our Strapi CMS, with clearly defined rights and without going through the clipboard. The process from content creation to go-live should thus become more effective and faster.

A Claude integration of this kind is not just an IT topic. It changes how editorial and development teams work together in the Headless CMS. The editorial team prompts what is needed, Claude takes care of the structure and preparation in the CMS, and people review the results before they go live and make final changes. For companies with multiple editors and many content types, this adds up quickly.

What Claude AI specifically does in the CMS

Claude is the AI assistant from Anthropic. In everyday life, many know it as a chat where you can ask questions or have texts written. There are several layers: Chat, Cowork, and Code. In an integration with the CMS, Claude Code additionally gains access to tools. Instead of just delivering text, Claude can perform actions in the system, such as reading content, preparing entries, or filling fields. The allowed actions are defined by you. In comparison, Claude Cowork has access to defined folders and can create documents but cannot make changes to systems.

The connection runs via the Model Context Protocol, or MCP for short. MCP is an open standard that describes how an AI model accesses external systems in a controlled manner. You can think of it as a defined interface: The CMS specifies what Claude can see and do, and Claude adheres to that.

For decision-makers, this is interesting for three reasons:

  • Less manual work: Claude takes over routine steps in the CMS. Formatting pages takes seconds instead of hours.
  • Control: Access is limited by rights, and people approve content. All pages are saved as drafts. They are published by the Human-in-the-loop.
  • No rebuilding: The existing CMS remains; it just gets an additional interface.

For those wanting to get an overview of the possibilities and limitations of AI in the company, you can find it in our article on the Opportunities and Risks of AI in Business..

How the Strapi MCP Server connects the CMS with Claude

For our Claude integration, we use the official Strapi MCP Server, which is directly integrated into Strapi 5. It connects our CMS with Claude Code, which we use via Claude Desktop.

The big advantage: Strapi already provides the necessary functions. We didn't have to develop our own interface and can let Claude work directly with our existing content types.

Requirements and setup

The requirement is Strapi 5.49.0 or newer, as the official MCP server has been generally available since this version. Additionally, you need Claude Code and the appropriate access permissions for Strapi. The Strapi MCP server is included for free in the Strapi subscription.

The setup is relatively simple: You don't need to activate the MCP server in the current versions of Strapi; it works out of the box. Previously, a code change was necessary for this. All that remains is to configure an API token for authentication and establish the connection in Claude Code. The steps are described in the Strapi documentation in the section on connecting with Claude Code.

The setup requires some basic technical understanding but can be done well even without in-depth AI knowledge. Once the connection is established, Claude AI can be operated using natural language without needing to write code yourself.

Permissions and security

The official Strapi MCP Server uses the existing permission system of Strapi. Access rights can control which content Claude can read, create, or edit.

When creating content with AI, it is important to us that control remains with the editorial team. Claude assists in preparing the content, while people review the results and decide on publication.

This way, we combine the advantages of the Claude integration with a controlled and traceable editorial process.

An example from the editorial daily routine

A typical example of content creation with AI is preparing a blog article. Instead of manually creating content in Strapi, we describe to Claude what topic should be covered and what structure the article needs.

Claude can then use the appropriate content type, fill in the designated fields, and prepare the article directly in the CMS.This includes, for example, titles, body text, and SEO metadata.

This form of automated content creation saves manual work steps. The editorial team reviews the finished draft and decides on publication.

Content creation with AI in the editorial daily routine

Content creation with AI initially sounds like texts at the push of a button. But in our setup, it’s so much more: Claude supports the process from the order to the finished draft in the CMS. The basic idea is simple. A person (for example, a developer) gives the order and provides the technical input, Claude prepares the content in the CMS, and a person (for example, a marketing employee) reviews and approves it. Without Teams messages, Google Docs, or screenshots being sent back and forth between departments.

This has a clear advantage over traditional chat. Because Claude works directly in the CMS, content lands in the right fields and components with the correct formatting and metadata. Because let’s be honest, how many of you reliably enter an alt text for every image or have copied all SEO data 100% correctly? Titles, descriptions, and other details no longer need to be transferred manually. This saves time and reduces the number of errors that occur during copying.

We emphasize the distribution of roles. Claude delivers drafts and handles tedious tasks. Content direction, technical accuracy, tone, and the decision on publication always remain with the editorial team.

Our Claude integration can be used for different content types, including articles, pages, and events. What content Claude can edit depends on the available content types and the assigned permissions.

Claude can not only create texts and assemble new pages based on existing patterns but also fill structured fields like titles, descriptions, slugs, and SEO metadata. This is particularly useful for more complex content structures with multiple components.

The content creation with AI is meant to support the editorial team, not replace it. Content is still reviewed by people before publication.

How high the Claude costs in the CMS really are

The question about the Claude costs cannot be answered with a single number, as they depend on usage. It makes sense to break down the costs into four components:

  • Claude usage: Depending on usage, there are costs for a subscription or for consumption via the interface. The current prices are set by Anthropic and can change, so it’s best to check directly with the provider.
  • Setup: Connecting the CMS via the MCP server is a one-time development effort but is quickly implemented since it only needs to be configured and connected.
  • Operation and maintenance: CMS, MCP server, and Claude continue to evolve. Updates and adjustments are part of ongoing operations.
  • Review by the editorial team: The time the team spends on review and approval also needs to be factored into the equation.

These costs are offset by saved work hours, especially for routine tasks that were previously done manually in the CMS. Whether the investment is worthwhile can be shown by a simple comparison: How much time does a task cost today, and how much time is left when Claude does the preliminary work?

To be honest: We did not collect specific numbers for this article. We describe the cost structure in a way that you can apply it to your own company. But just in formatting the already existing pages, this setup has already saved us quite a bit of work time. And it allows us a new, leaner workflow where the article is created directly in the CMS and approved for review.

Where automated content creation saves time and what else it brings

The biggest effect is seen where many small steps come together. Automated content creation rarely saves huge amounts of time at a single point; it saves a little at many points. Together, this creates noticeably more room for editorial work. The advantages at a glance:

  • Faster: Fewer steps between idea and draft in the CMS, no copying between tools.
  • Cheaper: Routine work takes less time and thus less money per content.
  • More consistent: Structure, fields, and metadata follow the same guidelines every time.
  • Fewer errors: Less manual transferring means fewer typos and forgotten fields.
  • More focus: The editorial team focuses on content and quality instead of maintenance work.
  • Scalable: When more content is needed, the effort grows slower than the amount.

These advantages only occur if the integration is set up properly and the team supports the process. AI in the CMS is a tool for collaboration among the editorial team (from different teams) and does not replace it. We explain in detail which processes are suitable for a start in our article on how to find suitable processes for AI. What we learned and where the limits of the Claude Strapi integration lie

AI in the CMS brings a lot but has limitations that you should plan for from the start:

Review remains mandatory:

  • Claude can make mistakes. Content needs a technical review before going live. Rights should be granted consciously:
  • The more Claude is allowed to do in the CMS, the more important clear boundaries become. The process needs maintenance:
  • If you create a new collection in the CMS or add fields to an existing collection, you need to enable these in the API key for the MCP server to access them. What particularly positively surprised us about the Claude Strapi integration was the easy setup.

Thanks to the official MCP server, no custom interface was necessary. . The operation via Claude Code is also straightforward: Tasks can be described directly in natural language.. Using Claude Code is also straightforward: tasks can be described directly in natural language.

The biggest challenge lies less in the connection itself, but in formulating clear instructions and controlling the results. Especially with complex content structures and nested components, careful review is important.

Our most important learning: Automated content creation doesn't have to be complicated. With the right permissions and clearly defined workflows, AI can be relatively easily integrated into existing editorial processes. The professional responsibility remains with humans.

Frequently asked questions about the Claude integration in Strapi

What is a Claude integration in the CMS?

With a Claude integration, the AI assistant Claude gets controlled access to your CMS. Claude can then read and prepare content without anyone needing to copy text between the chat window and the CMS. You define which actions are allowed through permissions.

What are the costs of Claude and the connection to Strapi?

The costs for Claude consist of usage at Anthropic, a one-time setup fee, ongoing operation, and the editorial review period. There is no flat rate as it depends on usage. Current prices are provided directly by Anthropic.

What is a Strapi MCP server?

The official Strapi MCP server allows AI assistants like Claude to communicate directly with a Strapi CMS. Through the Model Context Protocol, content can be read, created, and edited.

For our Claude integration, we use the MCP server integrated directly into Strapi 5. A separate interface is therefore not necessary.

Can Claude automatically publish content in the CMS?

Yes, the official Strapi MCP server generally supports the publication of content, provided the appropriate permissions are in place.

However, for automated content creation, we recommend having Claude prepare drafts first. The editorial team reviews the content and then takes over the publication. This way, control remains with humans.

If the content is not editorially reviewed by a human before publishing, the EU's labeling requirement for AI-generated content applies.

For which companies is Claude AI in the CMS worthwhile?

The use is worthwhile for companies where many similar tasks regularly occur in the CMS, such as creating and maintaining entries. The more routine work the team currently does manually, the greater the potential benefit.

In the AI workshop [KITS] of LEAN-CODERS, we work with you on three prioritized use cases and a roadmap, in a half-day session for a fixed price. If your headless CMS needs support, get in touch with us.

Do you want to use AI meaningfully in your company?

this is the end my friend

Who wrote it

Djordje Savanovic Fullstack Developer

Djordje develops frontends with a focus on user experience. As a Senior Developer, he accompanies projects from User Story to Release, works with React, React Native, and Next.js, optimizes systems, modernizes architectures, and consistently thinks in products.