What is Webflow MCP 2.0?

Webflow MCP is Webflow's Model Context Protocol server. It connects AI tools such as Claude, ChatGPT, and Cursor to your Webflow account so agents can design, build, and manage your sites from inside those tools. Version 2.0, launched July 21, 2026, adds full design-system support, CMS-connected dynamic content, custom font management, agent instructions, conversational site analytics, and governance features like branch-based building, permission enforcement, and activity logs.

Does Webflow MCP 2.0 cost extra?

No. Webflow states that MCP 2.0 is available to all users at no additional cost.

Do I still need the Webflow bridge app with MCP 2.0?

For most work, no — MCP 2.0 eliminates the bridge app for most use cases, which also makes it possible to build across multiple pages or sites at once. The bridge app is still required for tasks that need live visual context, such as snapshots or prompts that point at a specific element on screen.

Can AI agents break my live Webflow site through MCP?

Webflow added governance features to prevent exactly that. Agents can build on branches — an isolated environment — before changes are pushed live. The MCP enforces your account roles and permissions, including custom roles and per-site, per-collection, per-page, and per-locale access. Every MCP action is recorded in the site activity log with who triggered it and what changed.

Does Webflow MCP help my site's visitors?

Not directly. MCP works on the builder side: it helps you and your agents create and manage the site. Visitors browsing your published site do not interact with MCP. Answering their questions on the page — about services, pricing, availability, policies — is a separate, visitor-facing job that an on-site AI assistant handles.

Webflow update · July 2026

Webflow MCP 2.0: your site just became something AI agents can build

Webflow has spent the summer wiring itself into the AI-agent ecosystem, and MCP 2.0 is the centerpiece. Since July 21, 2026, agents running in Claude, ChatGPT, or Cursor can build pages with your design system, manage CMS content, analyze traffic, and answer for every change they make. If you own or manage a Webflow site, here is the plain-English version.

Launched July 21, 2026 · Available to all Webflow users at no additional cost · Works with Claude, ChatGPT, Cursor, and other MCP clients

What is Webflow MCP, in one paragraph?

MCP (Model Context Protocol) is the plug standard that lets AI assistants connect to outside tools. Webflow's MCP server is the socket on Webflow's side: once connected, an AI agent inside Claude, ChatGPT, or Cursor can see and operate your Webflow sites — the way it might otherwise operate a code editor. Version 1 could already do useful work; version 2.0 is the point where Webflow treats agents as first-class site builders.

What MCP 2.0 adds

Design & build

Agents use your real design system

Agents build with your Webflow styles, elements, and components — including props, slots, and variants — and connect them to CMS data so content stays dynamic. They can also register, upload, and manage custom fonts. The bridge app is gone for most use cases, so agents can work across multiple pages or sites at once.

Agent instructions

You tell agents how to work

Point agents at your design system, CMS content, legal documents, or voice-and-tone guidelines, and define skills and rules for how tasks should be done. The same instructions feed both Webflow's own AI and third-party agents connecting via MCP.

Analytics

Ask your site how it is doing

Agents can answer questions about traffic trends, top pages, engagement, time on page, and traffic sources in conversation, and pass that data to other MCP-connected tools like Amplitude, Slack, or Asana.

Governance

Guardrails for production sites

Agents build on branches before anything touches the live site. The MCP enforces your roles and permissions — down to site, collection, page, and locale — and every MCP action lands in the activity log with who did what, queryable via MCP itself.

All of it is included for every Webflow user at no additional cost, per Webflow's announcement.

The bigger picture: Webflow is going agent-native

MCP 2.0 did not arrive alone. Look at Webflow's release list for one month:

Jul 1, 2026Webflow app in ChatGPT — audit SEO, fix content, and update the CMS from inside ChatGPT.

Jul 21, 2026MCP 2.0 — agents build, manage, and analyze sites with governance.

Jul 28, 2026AEO content optimization agents — agents that write and optimize content to get cited by AI search.

Jul 30, 2026Perplexity in AEO analytics — track your brand's visibility across ChatGPT, Gemini, Claude, and Perplexity.

The direction is unmistakable: Webflow is betting that the people and programs working on your site will increasingly be AI agents, and that AI search engines will increasingly decide who sees it. For agencies and freelancers who manage client sites on Webflow, this is a genuine productivity shift — one operator with good agent instructions can now maintain a portfolio that used to need a team. Webflow is also far from alone in this race: our AI-agent platform tracker keeps score across Shopify, Wix, Shopware, and the rest.

What MCP does not touch: the visitor

Every capability above points in one direction — at you, the builder. Agents help you design, publish, analyze, and optimize. But when the site is live and a visitor lands on it with a question — Do you ship to Norway? Is the studio free on Thursday? What does the Pro plan include? — there is no agent on their side of the glass. They read, they hesitate, and if the answer is not on the page, they leave.

MCP 2.0 makes your Webflow site easier to build. It does nothing to make it easier to ask.

That visitor-facing half is what Asyntai's AI chatbot for Webflow covers: a chat assistant on your published site that answers visitor questions using your own content, in the visitor's language, around the clock. It installs with one snippet in Project Settings → Custom Code, so it works on any Webflow plan and any template. A site built by agents and answered by an assistant is covered on both sides of the glass.

Frequently asked questions

Which AI tools work with Webflow MCP 2.0?

Webflow names Claude, ChatGPT, and Cursor, and the MCP standard means other MCP-capable clients can connect too. Webflow also ships a dedicated ChatGPT app (since July 1, 2026) for site management inside ChatGPT.

Is MCP 2.0 free?

Yes. Webflow states MCP 2.0 is available to all users at no additional cost.

Can agents change my live site without review?

Not if you use the guardrails. Agents can work on branches in an isolated environment before updates are pushed live, permissions are enforced per site, collection, page, and locale, and the activity log records every MCP action with its author.

I run client sites on Webflow. What should I do with this?

Set up agent instructions once per client — design system, tone, rules — and routine maintenance becomes something you delegate to an agent and review on a branch. It compounds across a portfolio. Pairing it with a visitor-facing assistant on each client site is a natural upsell for the same reason: both turn ongoing work into configured-once systems.

Does any of this answer my visitors' questions?

No. MCP operates on the building side of your site. Visitors on the published site still need their questions answered on the page, which is what an on-site AI assistant like Asyntai does — it answers from your site's own content and captures leads while visitors are still there.

Agents build it. Who answers for it?

Webflow MCP 2.0 puts AI to work on your site's construction. Asyntai puts AI to work for your site's visitors — answering questions from your own content and capturing leads, live in minutes with one snippet.

Get Started Free See the Webflow integration