Learn how the Webflow MCP server connects AI tools to your sites and Workspaces.
The Webflow MCP (Model Context Protocol) server connects compatible AI tools to your sites and Workspaces, letting you use natural-language prompts to design, build, and manage content.
You can use the Webflow MCP server with most MCP-compatible AI tools, such as Claude, Cursor, Codex, and Slack.
Before you get started
The Webflow MCP server is available on all Site plans, including the free Starter plan. Only Workspace owners, Workspace admins, and Site managers can authorize access to a site through the MCP server.
What you can do with the Webflow MCP server
Connecting your AI tool to Webflow's MCP server lets you work with your site using natural-language prompts. Here are some examples of what a connected AI tool can help with.
Design and build
- Create and edit pages
- Build new layouts or change existing ones
- Manage styles and element settings
- Connect elements to CMS Collections
- Create, modify, and manage components
- Build and manage a design system
Manage your site content
- Create and manage CMS Collections and items
- Organize and optimize assets
- Manage and improve SEO and AEO
- Identify and fix common accessibility issues
- Read and add comments
- Publish your site or individual pages
Analyze site performance
These features require the Analyze add-on.
- Ask questions about your Webflow Analyze data
- Summarize traffic trends across date ranges
- View top pages, traffic sources, top events, and average time on page
- Turn performance insights into summaries, reports, or team updates
Learn more about asking AI about your Webflow Analyze data.
Good to know
Webflow Enterprise teams can use page branches and the Site Activity Log to manage AI-assisted work through the MCP server. Page branches let your AI tool build and edit in an isolated branch before changes are merged, and the Site Activity Log shows MCP activity with AI attribution so teams can review what changed.
Connect your AI tool to the Webflow MCP server
Many AI tools connect to the Webflow MCP server through a built-in connector or app, right inside the tool you already use. Follow the guide for your tool below to connect it to your Webflow sites and Workspaces:
If your AI tool isn't listed, you may still be able to connect to the Webflow MCP server manually.
Important
When you connect your AI tool to the MCP server, you can choose which sites or Workspaces it can access. We recommend that you only authorize sites you're actively working on to maximize security and performance.
Guide your AI tools with instructions
Instructions let you store reusable context with your site, such as brand voice, content structure, design patterns, component usage, or team preferences. The Webflow AI Assistant and third-party AI tools connected through the MCP server can reference those instructions, so you don't need to repeat the same guidance each session.
Instructions include skills and rules. A skill is detailed guidance for a particular kind of work or task — like writing product copy, creating campaign pages, or using your components — that a tool can reference when it's relevant. A rule is a short, fixed instruction that applies to everything a tool does — for example, always writing headings in sentence case.
You can create and manage instructions in Webflow from the Instructions panel in the left toolbar, work with them through the MCP server, and share them across sites with Shared Libraries. You can also add references to your site's resources, such as components, variables, CMS items, pages, assets, and other instructions.
When an agent works in your site, it should read your rules before taking action and reference the relevant skills for your request — and you can point it to a specific skill by naming it in your prompt. How reliably a third-party AI tool applies your instructions can vary, so when you start a new session, it's a good idea to ask the tool to read your site's instructions first. You can also export your instructions and add them to your AI tool directly.
Learn more about creating and managing instructions.
Current limitations
The Webflow MCP server can help with many design, content, and site-management tasks, but it can't complete every action available in Webflow yet. Some tasks may still need to be completed directly in Webflow. Current limitations include:
-
Localized content: It can read and update existing localized CMS items and static content in secondary locales, but can't create new localized CMS items.
-
CMS bindings: It can't control some binding settings, including conditionals, date or number formatting, and legacy bindings.
-
Collection lists: It can't set the source, filter, or sort options.
-
Element settings: It can't control every setting for every element type.
-
Interactions and animations: It can't create or edit them.
-
Canvas-based actions: It needs the Webflow MCP Bridge App for actions that depend on what you have selected or open on the Webflow canvas.
-
Analyze data: It can read site-level performance only — page-level analysis, goal reporting, and filtered queries aren't available yet.
FAQs
How does the Webflow MCP server work?
The Webflow MCP server acts like an interpreter between your AI tool and Webflow. When you ask your AI tool to take an action, the MCP server translates that request into the Webflow API calls needed to complete the task, runs those calls on your site, and returns the result to your AI tool.
Do I need to have my site open in Webflow?
Most AI tools can use the MCP server to take actions on your site directly, so you don't need to have your site open in Webflow. However, certain actions that need page-specific context may require you to open your site in Webflow and launch the Webflow MCP Bridge App.
What plan do I need to use the MCP server?
The Webflow MCP server is available on all Site plans, including the free Starter plan.
Who can authorize access to a site through the MCP server?
Only Workspace owners, Workspace admins, and Site managers can authorize access to a site through the MCP server.
Which AI tools work with the MCP server?
MCP is an open standard. If your AI tool supports MCP, you can connect it to the Webflow MCP server. Popular tools — such as Claude, Cursor, Slack, and Codex — work with the MCP server, along with many more.
Can I limit which sites my AI tool can access?
Yes — when you first connect your AI tool to the MCP server, you can choose which sites or Workspaces it can access.
Does the MCP server use Webflow AI credits?
No — using the MCP server doesn't consume Webflow AI credits. Instead, usage is counted toward your AI tool's plan or credits.
Learn more
Help Center articles:
Developer documentation: