What is the Model Context Protocol (MCP) and what does it mean for your CMS and website?

In the heavy media coverage of everything to do with AI, you will have undoubtedly seen the Model Context Protocol (MCP) being mentioned a few times, including on this blog. This emerging standard is opening up possibilities to connect your CMS or DXP platform to LLMs and AI tools and services, supporting interoperability: unsurprisingly major platforms including Umbraco, Sitecore and Optimizely have already moved to support using MCP.

It may be that you already know about the MCP and what it does, but if you don’t, then this blog is for you. We’re going to do an MCP 101, explore how the standard can impact your CMS or website, and then see what some of the major players are doing in this area.

What is the MCP?

The Model Context Protocol is a standard which helps connect LLMs and related AI services such as Claude and ChatGPT to external tools and data sources, enabling integrations and the exchange of data. In a nutshell, it seeks to standardise the way AI tools and external systems such as your CMS talk to each other, ultimately enabling connected experiences.

The MCP was originally created by Anthropic in November 2024 but is now open source so everyone across the industry can use it. Because other major AI providers beyond Anthropic such as OpenAI have embraced the use of the MCP, it means that that it has effectively emerged as the industry standard. With so many organisations wanting to connect AI tools to different external systems, it means that software providers – including CMS vendors – are also embracing the use of the MCP standard.

What problem does the MCP solve?

A CMS often includes integrations with different platforms and products; for example, an e-commerce solution, an analytics engine, a personalisation tool, a security product, a CRM system, a headless set-up and so on. Of course, the latest tools that organisations want to integrate with their CMS are AI tools like Claude.

Over the years, integrations have been made much easier through out-of-the-box connectors and also APIs, but it can still take considerable effort. An integration using an API will involve developers, can be complex, add costs, needs to be maintained and slows down implementation. That can be a problem, particularly given the pace of change involving AI and the desire to integrate AI into different systems, including your CMS.

APIs are still critical, but the MCP standard works on top of APIs, so any new integration with an MCP-compatible system should “just work” and be easier, faster and cheaper to execute.

What does the MCP mean for your CMS or DXP?

The MCP brings several benefits to your CMS platform.

Easier to connect your AI of choice

As long as your CMS supports it, the MCP significantly reduces the barriers to connect and integrate your AI of choice into your CMS. This reduces costs and potential reliance on external IT resources (although of course we’re always happy to help!), and also any delays in a project.

Navigate changes in AI tools

As the MCP is a standard it means it is far easier to switch AI tools and assistants, or add additional LLMs to integrate with your CMS, without disruption. With the evolution of AI incredibly fast, it means you’re not held back by any decisions to switch, add or upgrade the AI you choose to help power your website.

Accelerate AI innovation with your CMS vendor

The MCP is helping to accelerate the development of innovative new features within your CMS. This is particularly true of Umbraco who follow more of a “bring your own AI” policy that encourages customers to plug in their own LLM and AI assistant, rather than develop AI-native features.

Bypass the need for AI-native features

Your CMS vendor may be introducing some AI features into the CMS which are native to the platform. However, these might not be what you need, might not work in the way you want, and may incur an additional cost. Being able to plug your own AI via the MCP can help you bypass reliance on AI-native features and maximise your investment in your LLM subscriptions.

Maintain governance and security models

The MCP standard preserves governance and security models, for example maintaining any permissions that control access. You may also need to use a particular AI model or LLM because it ticks boxes around data residency, privacy and security, so one advantage of the MCP is that it should allow you to easily maintain any governance in place.

Supports the use of AI agents

The age of agentic AI is upon us and although organisations are only dipping their toe into using agents (and many AI features are now being rebranded as agents), this all looks set to scale pretty quickly. The MCP standard opens up the opportunity to deploy AI agents across different systems and products that will have more value.

Work from your AI assistant, not just your CMS

In terms of actually building AI into your CMS and editing experience, the MCP offers two options. It can bring conversational prompts and digital assistant directly into your CMS experience that can be used by admins and content editors to create content, instruct AI features and prompts supported by your CMS relating to a wide variety of areas from analytics to personalisation, and more. AI can bring a lot of power to your content creation, optimisation, analytics, and more.

However, while you might want to add AI to your admin and editing interface, you may also want to use the MCP to be able to instruct your CMS directly from ChatGPT or Claude, if that’s where you do a lot of your work.

Note that not every CMS supports all of these capabilities yet, but they are typical of the direction of travel.

Are there any limitations with the MCP?

Like everything with AI, things are moving incredibly fast. In the rush to be MCP-ready, some CMS features might be not fully ready, and effectively be in Beta, so not everything may work perfectly or be as stable as it should be.

Security and data privacy are critical. So, while the MCP should preserve areas such as permissions, connecting to any third party must always be considered very carefully and reviewed by your security team. You need to know that all data privacy, residency and security boxes are ticked by the solution you’re connecting to.

In theory, also the MCP should make integrations much easier, but digital marketing teams may still need to rely on IT resources to make changes.

Using natural language prompts to carry out changes which you may have previously used a CMS interface, or a WYSIWYG editor can be a significant learning curve, and it might take some time to master, or even not be for particularly people.

Finally, using AI for website management has upsides and downsides. Introducing more AI into your CMS may open up greater opportunities, but AI-created content can produce errors, be off-brand and contain mistakes, so how you use it needs to be carefully considered.

How are the major CMS providers supporting the MCP?

The major CMS providers are enabling digital marketing teams to take advantage of the MCP in different ways and at different speeds and these are likely to evolve going forwards. Here’s the situation with three CMS products 3chillies works with: ·

• Sitecore has launched the Marketer MCP which allows users to connect digital assistants like Claude directly to Sitecore. This is included in all subscriptions and is designed for a non-technical audience.
• Optimizely recently launched the Remote MCP Server for the experimentation features of Optimizely, allowing you to work directly with the platform from your AI assistant. Official remote MCP support for the main CMS and for Optimizely’s Analytics product are on the roadmap.
• Umbraco has moved quickly to support MCP, with an emphasis on using the standard to integrate AI tools and LLMs with Umbraco rather than building native features within the platform. The centre of the MCP offering is the Umbraco MCP Server which enables developers to connect LLMs, but since then there has also been the announcement of a number of “agent resources” including:
  • Tools and templates to enable MCP implementations
  • Agent Skills, to help LLMs use the right information
  • A new Editor MCP that will bring AI capabilities to content and editorial teams without IT involvement.
  • There may be further announcement soon at the upcoming Codegarden conference.

Watch this space

The MCP standard is already here with the potential to integrate AI assistants into your CMS admin and editing experience. This is moving from a stage where it was something for developers to think about, to something which marketers need to know about.

As with anything relating to AI, things move quickly so expect both the standard and support across each CMS product to evolve and improve.

If you’d like to discuss how you can take advantage of the MCP to improve your CMS or website, then get in touch!