The remote server connects ChatGPT, Claude, Cursor, Grok and other assistants directly to Framer projects, with OAuth sign-in, account-level permissions and an optional branch-based review step before anything goes live.
Framer has released Framer MCP, an official Model Context Protocol server that lets AI agents outside its own editor work inside Framer website projects. Through it, an agent can create pages, edit layouts and copy, manage CMS collections and review SEO, and every change stays editable in the Framer canvas afterward.
The server is hosted at https://mcp.framer.com/mcp and works with any AI client that supports remote Streamable HTTP MCP servers and OAuth. Framer has published one-click listings for ChatGPT and Codex, Claude and Claude Code, Cursor and Grok Bot, and setup guides for Grok, OpenCode, VS Code, Kiro, Notion, Pi, Google Antigravity, Windsurf, Warp and Devin.
For enterprise marketing teams, the practical change is where website work can start. A marketer already working in ChatGPT, Claude or Notion can tell that assistant to update a landing page or load a batch of blog posts into the CMS, without opening Framer or handing the request to a web team. The Framer editor does not need to be open while the agent works.
What the server can do
Framer’s documentation describes five main areas of use:
- Build and edit pages. Agents can create new pages or change existing layouts across desktop, tablet and phone breakpoints, reusing the project’s existing styles and components.
- Manage content and the CMS. Agents can create collections and items, update structured content, and import content from sources the AI client can reach.
- Review SEO and site quality. Agents can inspect metadata, headings, alt text and page structure. Framer notes that if a client is also connected to an SEO tool such as Ahrefs, the agent can combine that data with the Framer project during its review.
- Work with outside tools and data. Agents can bring files, scripts, APIs and research available in their client into the Framer project. What is available depends on the client and what the user has connected to it.
- Create a branch for review. Agents can make changes on a Framer branch so the work can be checked before it is merged or published.
The cross-tool point carries the most weight for marketing operations. The MCP server is one connection among several a team may already have in its AI client, so a single request can pull keyword data from an SEO platform, draft copy, and write the resulting pages into Framer.
Setup and access
Users connect by installing Framer from a client’s plugin or connector directory, or by adding the server URL manually. Either way, Framer opens a browser window where the user signs in and approves the connection. The agent then operates with the permissions of the Framer account that signed in, and can reach the canvas, components, supported project settings and CMS in any project that account can access.
Framer’s troubleshooting guidance tells users to check with their workspace administrator if plugins or MCP connections require approval, which indicates some organizations will have admin-level controls over whether these connections are allowed.
AI usage is billed through the agent, not Framer. According to Framer’s FAQ, usage follows the plan or credits of whichever AI tool the user connects. Framer’s published materials reviewed for this article did not state whether Framer MCP access itself requires a particular Framer plan.
Publishing and review controls
By default, an agent’s changes stay in the project and do not reach the live site until someone publishes. Framer confirms an agent can publish if the user explicitly asks it to. For review, Framer recommends asking the agent to work on a branch, then comparing that branch with main before merging or publishing.
This differs from Framer’s earlier External Agents setup, a local command-line bridge (@framer/agent) that Framer now labels as legacy. Framer’s documentation for that setup says every change an external agent makes automatically happens on a branch. In the MCP documentation, branching is described as something the user asks the agent to do. Teams moving from the local setup to MCP may want to confirm how branch-based review is handled in their own workflows.
What’s not included yet
Localization is not supported in Framer MCP at launch. Framer directs teams with localization workflows to the legacy local CLI setup, and says both connection methods will remain supported until Framer MCP reaches feature parity.
Context
The launch replaces several workarounds. Before today, connecting outside AI tools to Framer meant using Framer’s local External Agents bridge, a community-built MCP plugin on the Framer Marketplace, or third-party MCP services built on Framer’s Server API, which entered open beta in February 2026. As recently as mid-September, a Framer help article told users they did not need a separate Framer MCP server. Framer’s External Agents page now calls Framer MCP the recommended way to connect AI tools.
Framer is also inviting AI product companies to integrate. Its MCP page includes a submission option for developers who want their own MCP client to let users create, edit and manage Framer projects from inside their product.
The release follows a run of agent-focused updates from Framer, including Skills on September 22, which lets teams save reusable instructions for the built-in Framer Agent covering design systems, writing style and CMS workflows.
Considerations for enterprise marketing leaders
- Permission scope. Because the agent inherits the signed-in account’s permissions across every project that account can access, teams may want to consider which accounts are used to authorize agent connections, particularly where one login spans multiple brands or client sites.
- Publishing guardrails. Agents can publish when asked. Organizations with formal approval steps for web changes may want to decide whether branch-based review should be a required practice for agent work.
- Admin controls. Framer references workspace-level approval for MCP connections. IT and marketing operations teams may want to confirm what controls are available on their plan before broad rollout.
- Cost visibility. With AI usage billed through each user’s chosen assistant, spend tied to website work may show up across several AI vendor accounts rather than in Framer.
Framer MCP is available now. Setup instructions are at framer.com/help/articles/getting-started-with-framer-mcp.
