Working Smarter with Your DXP: The Case for MCP-Powered AI Operations

There's a question that every digital team faces when they begin exploring AI capabilities within their platform: how do you move from using AI as a conversation partner to using it as a genuine operational participant? 

The answer, increasingly, is the Model Context Protocol, or MCP. And the platforms that have embraced it most seriously are beginning to demonstrate what AI-native content operations actually look like in practice. 

What MCP is, and why it matters 

MCP is an open standard developed by Anthropic to standardise the way AI systems interact with external tools, data and services. Think of it as a universal interface layer, described by Contentful's team as "the USB-C of AI," that allows any AI agent to discover and interact with a system's capabilities without requiring a bespoke integration for each use case. 

At a technical level, MCP defines how AI agents can discover available tools, call operations, receive structured responses, and chain those operations together into workflows. Without MCP, connecting an AI assistant to your CMS requires custom integration work for each AI tool you want to support. With MCP, any compatible agent can connect, discover what your system can do, and start taking action. 

For digital teams, the practical implication is significant. MCP enables AI to move from answering questions about your content to actually managing it: creating entries, updating metadata, running bulk operations, triggering workflows, publishing assets, all using natural language as the instruction layer. 

How Contentful has implemented MCP 

Contentful has invested seriously in MCP, offering both an open-source local server and a remote hosted MCP server at mcp.contentful.com. The official server is OAuth-authenticated and scoped per space and environment, meaning teams can grant AI agents access to specific parts of their content infrastructure without opening up the entire system. 

Through the Contentful MCP server, AI agents can perform the full range of content management operations: creating, updating and publishing entries and assets; exploring and modifying content types; managing environments and spaces; running bulk tagging operations; and executing migrations. The key capability this unlocks is not individual task automation — it's operational scale. 

Consider what this means for a content team managing a large, multi-brand digital estate. Content audits that previously required analysts to manually review hundreds of entries can be run by an agent in minutes. Metadata inconsistencies across thousands of assets can be identified and corrected in a single workflow. Localisation tasks — translating and publishing content variations across multiple locales — can be orchestrated by agents rather than executed manually. 

Contentful has open-sourced the server deliberately, reflecting the view that the future of AI in content operations should be transparent and auditable. Organisations can inspect and customise the server to fit their specific needs. Early adoption data shows particular traction in media, retail and technology sectors, with teams using MCP for operational automation, migration acceleration and content governance at scale. 

How Sitecore has implemented MCP 

Sitecore's approach to MCP is arguably its most technically ambitious architectural decision in recent years. Within SitecoreAI, every product in the stack is defined as an MCP action. This means that any agent, operating within any workflow, can interact with any part of the Sitecore ecosystem, including XM Cloud content, CDP customer data, Personalize rules, Search configuration and DAM assets, through a single, standardised protocol layer. 

This is what makes Sitecore's Agentic Studio genuinely cross-platform rather than just multi-product. When an Agentic Flow orchestrates a campaign from brief through to personalised publication, it does so by calling MCP actions across multiple products in sequence. The protocol is the connective tissue. 

For developers and architects, Sitecore's MCP layer also makes the platform extensible to external AI tools. An agent built outside the Sitecore ecosystem, in Claude, in Cursor, or in a bespoke enterprise AI tool, can connect to the Sitecore MCP layer and take action within the platform. The community MCP server for Sitecore, built on the GraphQL and Item Service APIs, supports content creation, publishing, workflow management and item operations across XM, XP and XM Cloud environments. 

What MCP-powered operations look like in practice 

The operational patterns that MCP enables are worth being concrete about, because the abstract promise of "AI-powered content operations" can obscure what's actually achievable today. 

For a Contentful-based digital team, a practical MCP workflow might look like this: a developer uses Claude Desktop with the Contentful MCP server configured. They instruct the agent to review all product content items updated more than six months ago and flag those with missing SEO metadata. The agent queries the space, retrieves the relevant entries, evaluates them against the metadata schema, and returns a prioritised list, ready for review without a single manual query. 

For a Sitecore-based team, the equivalent might be an Agentic Flow that monitors campaign performance signals daily, identifies underperforming content, generates personalised variants using Contextually Aware Content Agents, submits them for editorial approval, and publishes approved variants to the relevant audience segments. A workflow that previously required coordination across multiple tools and team members, now running with minimal human involvement. 

The governance consideration 

MCP-powered AI operations increase the scope and speed of what AI can do within a platform. That makes governance more important, not less. Both Sitecore and Contentful have built access control and permission scoping into their MCP implementations. Agents can only operate within the environments and spaces they've been granted access to, and audit trails record what actions were taken and when. 

For digital teams thinking about adopting MCP-based operations, the recommended approach is to start with contained, reversible workflows: content auditing, metadata enrichment, bulk tagging. Build confidence in how agents behave within the platform before moving to publishing and deletion operations. The technology is ready. The governance framework should be designed alongside it. 

MCP is not a future capability. For teams on Sitecore and Contentful, it is available today. The question is not whether to adopt it, but how to design the operational processes that make the most of it.