The Knowledge MCP: Our Platform Knowledge, Readable by Any AI System

A few days ago we made the case that an AI coding agent can scaffold a storefront in an hour, but that operating the result is the actual discipline. Here is the other half of that story, and this time it is about us. Laioutr now serves its complete platform knowledge over an open MCP endpoint. Any AI system, whether Claude, Gemini, GPT, or your own model, can read our documentation, our component and section schemas, field types, and conventions, machine-readable, before it builds a single component.

We call it the Knowledge MCP. It is the same idea we keep coming back to: a model that can generate code is only as good as the knowledge it has to hand. The question was never whether a model can build against Laioutr. It was whether it knows how. The Knowledge MCP closes that gap, and it does it on an open standard rather than a proprietary one-off.

Why an open endpoint, not a closed integration

Instead of maintaining a separate knowledge source for every model vendor, MCP gives every compatible client one shared language to pull the same documentation. Laioutr builds on exactly that standard instead of a closed in-house solution. And it is now official: Laioutr is listed in the Anthropic Connector Store, so a Claude client can connect to the endpoint directly, without a proprietary SDK.

What the Knowledge MCP actually serves

An agent does not just get the docs. It gets context, machine-readable, before anything is built:

• Search the documentation and pull the exact slice a build needs.
• Fetch component and section schemas, slot rules, and field types, so an agent knows which fields a section has and which blocks belong in which slot, instead of guessing.
• Look up conventions and concrete examples to build components, apps, and integrations schema-conform.
• Connect any MCP-capable model, so there is no lock-in to a single vendor, and you can switch or combine models without rebuilding access to the knowledge.
• Serve every request through clearly assigned rights, so you decide which projects and which parts of the documentation an agent may see at all.

The same thesis, one layer deeper

This follows the same principle as our frontend-as-a-service position: the platform should give humans and AI agents what they need to build correctly, then stay out of the way. The Knowledge MCP changes nothing about your frontend. It hands over the knowledge required to build against the platform, and it deliberately does not replace documentation or responsibility. The knowledge becomes retrievable; what gets built, and what gets shipped, stays your call.

MCP is also just one channel among several. The same platform knowledge is reachable from a ChatGPT plugin and a Claude Connector too, same knowledge, same open language. For enterprise commerce teams, the takeaway is simple: the backend stays stable, the frontend layer is built jointly by humans and AI agents, and every one of those agents now works from the documented state instead of a guess.

More on the platform: Agentic Frontend Management Platform, Frontend as a Service, and Composability & Orchestration. Explore the endpoint itself on the Knowledge MCP page.

Related reading: MCP for Commerce Frontends: How Agents Write to the Platform Safely and WebMCP vs. MCP for Commerce: Where the Agent Boundary Belongs.