MCP can look straightforward in a demo: an agent creates a CMS collection, updates page metadata, audits a site, or builds a section from a natural language prompt.

Trying it on your own site is different. You need to know how to get connected, what MCP can actually help with, and what to ask it to do first.

Who this course is for

This course is for designers and developers who know their way around Webflow and want to start using MCP in real work.

You should be comfortable with core Webflow features like Pages, CMS collections, Components, Styles, and site structure. You do not need prior experience working with AI agents.

If you're a marketer, strategist, or team lead trying to understand what MCP can do, you can follow along too. Just know that some actions depend on your Webflow role and permissions.

What you'll need

This course uses Claude for the demos. Claude is one of several agents that support the Webflow MCP. We'll use it throughout the course so the setup, examples, and prompts stay consistent.

Most of the concepts you'll learn here, including how to write effective prompts and how to prepare your site for working with an agent, apply across agents.

To follow along exactly with the demos, you'll need a Claude account with connector support Check your plan if you're not sure whether connectors are available to you.

You'll also need access to a Webflow site where you can safely practice. A test site, cloned site, or non-production site is best while you're learning.

Only Webflow site owners and admins can authorize the MCP server. If you don't have that level of access, ask a site owner or admin to complete the authorization step before you begin. We'll look closer at that step later in this course.

What this course covers

This course takes you from "I've heard of MCP" to "I can run a real MCP workflow and understand what happened."

You'll learn how MCP connects an AI agent to Webflow, how to think about the work an agent can do, and how to write prompts that give the agent enough context and boundaries to be useful.

This course covers four main areas:

• How MCP works: what MCP is, what an agent does, and how MCP connects to Webflow
• How to get connected: how to authorize the connection and understand what that access means
• How to prompt for stronger results: how to give context, set constraints, and ask the agent to share a plan before acting
• How to keep going: how to check the result, avoid common first-session mistakes, and try a guided MCP activity on your own site

You'll also see both major ways MCP can work with a Webflow site: through site data, like CMS collections and page metadata, and through the Webflow canvas, like elements, components, styles, variables, and sections.

By the end, you'll have a practical starting point for using MCP in Webflow, plus a guided activity you can run on your own site.

Ready to continue?

Click Complete & continue to learn what MCP is and how it works before you connect anything.