Webflow's official MCP server connects an AI client to the Webflow Data API — CMS collections and items, site and page data — so Claude can populate a blog, bulk-edit CMS entries, or audit your content. Here's the setup.
Option A — official remote server
Add https://mcp.webflow.com/mcp as a remote/OAuth connector and authorise through Webflow:
{
"mcpServers": {
"webflow": {
"url": "https://mcp.webflow.com/mcp"
}
}
}
Option B — local via npx
{
"mcpServers": {
"webflow": {
"command": "npx",
"args": ["-y", "webflow-mcp-server"],
"env": { "WEBFLOW_TOKEN": "your_api_token" }
}
}
}
Get an API token from Webflow → Site settings → Apps & integrations → API access. Restart the client; see MCP config file location.
Verify
Ask: "List my Webflow CMS collections and add a blog post item titled 'Hello MCP'." Real CMS data back confirms it works.
Common problems
unauthorized— wrong token or OAuth site; reconnect.- Can read but not write — the token/role lacks CMS write scope.
- No tools in client — Node 18+ for
npx; see Cursor MCP not working.
Going further
Webflow pairs with the Figma and Magic servers for a design-to-site flow, and a web-search server for content research. Browse the content category or curated loadouts. Keep tokens private — see MCP security best practices.