MCP Server

How it works

The Webflow MCP server implements Anthropic’s Model Context Protocol specification to standardize communication between AI agents and Webflow’s APIs. This allows you to interact with your Webflow projects using natural language in any MCP-compatible AI tool.

Architecture

The MCP server acts as a translation layer between AI agents and Webflow’s APIs. When you prompt your AI agent, the server:

  1. Receives the request from your AI tool (Claude Desktop, Cursor, etc.)
  2. Translates the intent into specific Webflow API calls
  3. Executes operations on your sites using OAuth-authenticated access
  4. Returns results back to your AI agent in a structured format

Remote deployment

The server runs remotely at https://mcp.webflow.com/mcp to enable OAuth authentication. This approach provides several benefits:

  • No local credentials: Authorize multiple Webflow sites without storing API keys on your machine
  • Secure access: Token-based authentication with automatic refresh
  • Easy updates: Server improvements deploy automatically without reinstalling

Webflow MCP Bridge App

The MCP server executes Designer API calls through the MCP Bridge App, which automatically installs to your authorized sites after OAuth authorization. This App must remain open in the Webflow Designer for agents to have access to Designer API tools. However, you can minimize it after it connects to the MCP server.

MCP Bridge App

This App is not required for calls to the Data API.

MCP resources

In addition to tools, the Webflow MCP server exposes MCP resources. Resources are read-only content that supported clients can load directly into context without calling a tool.

The Webflow Guide tool, which describes how to use Webflow’s APIs and the MCP server, is available as a resource to agents. In clients that support resource references, such as Claude and Cursor, you can @-reference the Webflow Guide in a prompt to pull it into context before the agent starts working.

Mode awareness

The MCP server reports the current Webflow Designer mode in tool responses so that agents can adjust their behavior to what the active mode allows. Tools that are not available in the current mode return a ModeForbidden error. Tool descriptions include the modes in which each tool can run.