Ploy MCP: how to edit your site from Claude, ChatGPT or Cursor, step by step

By Ranjit Bhinge, founder of Blur Studio · Updated · 9 min read

Short answer: Ploy is an AI website and marketing platform, and its MCP server lets an outside AI app, such as Claude or ChatGPT (AI assistants) or Cursor (an AI code editor), work on your Ploy site. MCP (Model Context Protocol) is a standard way for an AI app to use another tool. As of November 2026 you add the address https://ploy.ai/api/mcp as a custom connector, sign in to Ploy with OAuth (a sign-in that needs no password sharing or API key) and choose permissions. Start with read-only access, then add write and publish only when you trust the results.

Ranjit Bhinge

You want to change your website by chatting with Claude, ChatGPT or Cursor, and you want to know the exact steps, the permissions and what could go wrong.

I'm Ranjit, and I run Blur Studio: 15+ years building software, 120 apps taken live, and clients who have gone on to raise $10M+. I build AI systems and processes that help businesses do a lot more with a lot less time. I also work differently from most of what else is out there, more on that further down this page. Say hello.

Connecting takes a few minutes. In your AI app, add a custom connector named Ploy with the address https://ploy.ai/api/mcp, sign in to Ploy when a window opens, pick which workspaces and permissions the app gets, and approve. Ploy's docs say you need no API key or token, because each AI app signs in on its own. The harder part is deciding what to allow and how to check the work.

Ploy is a platform from Ploy, Inc. that builds, edits and publishes websites through a chat with its own AI agent. The MCP connection is included with every account, per its docs. This guide gives the steps for each app, explains the four permissions, lists the plan limits that block people, covers safety, and ends with a test plan. For background on the standard itself, read what MCP is and why a website owner should care.

This page may contain affiliate links from which Blur Studio earns commissions.

What you can do once connected

Per Ploy's MCP docs (last updated November 4, 2026), your AI app can:

  • Start and continue chats with Ploy's agent. The agent is the one that builds and edits the site, so most work still happens inside Ploy.
  • Read your sites, forms, documents, company visitors, conversations and artifacts.
  • Create, edit and delete databases, tables, columns and rows (see Ploy's database and CMS explained).
  • Publish your site to production.
  • Move a site to Ploy from a web address, or start a new one. You can also give Ploy a file by public link or attachment, up to 10 MB.
  • Show a screenshot of a page, check if a site is live and at which address, list integrations and create a workspace.

It also answers "how do I" questions from Ploy's docs. In Claude, Ploy adds three starters to the plus menu: Build my website, Move my site to Ploy, and Check Ploy's progress. Long jobs, such as building a site or moving one, can take from a few minutes to about an hour. Your AI app does not wait. It offers to check progress every few minutes, and closing the app does not stop the work.

It cannot do everything. The docs say the MCP server does not use API tokens, and your AI app can only do what the permissions you approve allow. Money-moving actions need your approval in Ploy itself (see the safety section).

Before you start: plan limits that trip people up

AppWhat Ploy's docs say
Claude (Anthropic's AI assistant) on its Free planOne custom connector allowed
Claude Team and EnterpriseAn Owner adds the connector first in Organization settings, then each member selects Connect
ChatGPT (OpenAI's AI assistant), web onlyCustom MCP servers are not available in the mobile apps
ChatGPT Business, Enterprise and EduFull access, including changes to data. On Business, only admins and owners can add a custom server for the workspace
ChatGPT ProPloy can only read data
Cursor, Claude Code (Anthropic's command line coding tool), Codex (OpenAI's coding agent)No plan note for the connection itself. Check your own subscription

You also need a Ploy account and, for edits that cost AI credits, enough credits. Work started through MCP spends the same credits as work done inside Ploy. The Free plan gives 500 credits a day, up to 2,000 in total. See Ploy's free plan for what that covers. If credits run out, Ploy pauses new work and tells your AI app when credits refresh. Work already done stays saved.

The four permissions, in plain words

After sign-in you choose All my workspaces in an organization, or One workspace. Then you keep or remove these permissions:

PermissionWhat it allows, per Ploy's docsRisk level
Chat with agentsStart and manage chats. Ploy's agent can change data, use credits and use connected servicesHigher than the name suggests
Read workspace dataRead sites, forms, documents, company visitors, conversations and artifactsMedium: it can see customer details in forms and visitor data
Write workspace dataCreate, edit and delete databases, tables, columns and rows. Needs ReadHigh: deletes are possible
Publish sites to productionPublish saved site changes to your live site. Needs ReadHighest: changes go public

Notice that "Chat with agents" is not harmless. Because Ploy's own agent can change things when your AI app talks to it, the chat permission alone can lead to edits and credit spend. The docs say that you can only change workspaces or permissions later by removing Ploy from the AI app and connecting again.

Step by step for each app

Menu names change often. If a label differs, look for the option to add a custom connector or custom MCP server.

Claude Desktop and claude.ai

  1. Open Settings, then Connectors (on claude.ai, Customize, then Connectors).
  2. Select Add custom connector. Name it Ploy and enter https://ploy.ai/api/mcp. Leave the advanced settings empty.
  3. Select Add, then Connect. Sign in to Ploy in the window that opens and approve access.
  4. Start a new chat, select the plus button, then Connectors, and make sure Ploy is on.
  5. Ask: "Please list my Ploy workspaces." If it answers, you are connected.

A connector added in Claude Desktop also works on claude.ai and in the Claude mobile apps. To share files from claude.ai, add ploy.ai under Additional allowed domains in Settings, then Capabilities.

Claude Code

  1. Run claude mcp add --transport http ploy https://ploy.ai/api/mcp.
  2. Start Claude Code, type /mcp, select ploy and choose Authenticate.

Claude Code can also upload a file from your computer in one step with a one-time link that expires after 10 minutes, per Ploy's docs.

ChatGPT (web only)

  1. Go to chatgpt.com/plugins and select the plus button, then Add custom MCP server.
  2. Name it Ploy. Enter the address as the Server URL. For authentication choose OAuth and leave client ID and secret empty.
  3. Read the warning, choose I understand and want to continue, then Create as a plugin.
  4. Install Ploy, sign in and approve access.
  5. In a new chat, type @ and select Ploy.

ChatGPT asks you to confirm before Ploy changes anything, per Ploy's docs. OpenAI's own page says it does not develop or verify custom MCP servers.

Cursor

  1. Open ~/.cursor/mcp.json and add { "mcpServers": { "ploy": { "url": "https://ploy.ai/api/mcp" } } }, merged with any servers you already have.
  2. Open Cursor Settings, find ploy in the MCP list and select Connect to sign in.

Codex and other apps

For the Codex command line tool, run codex mcp add ploy --url https://ploy.ai/api/mcp. Ploy's docs also list steps for the Codex desktop app, Grok (xAI's AI assistant), Grok Bot and Meta Muse (Meta's AI assistant). Any app that supports remote MCP servers with OAuth can connect: name Ploy, the URL above, transport Streamable HTTP, authentication OAuth.

First prompts to try

  • "Please list my Ploy workspaces." (If you chose One workspace, ask "Please list my Ploy sites.")
  • "Ask Ploy what to do next in this workspace."
  • "Is my site published, and at which address?"
  • "Show me this week's form submissions." (read-only)

Safety: what to set up before you let an AI edit a live site

An AI app that can edit your site can also make a wrong edit quickly. Here is what the vendors say, and the habits I would add.

  • Trust the server. Claude's help page says to connect only to remote servers you trust. OpenAI's page says it does not verify custom servers and warns about prompt injection, where hidden text in data tricks an AI into following it. Cursor's docs say to review permissions and use restricted keys. Ploy runs this server itself, so you are trusting Ploy and your AI app together.
  • Keep a person in the loop. The MCP specification says there should always be a human in the loop with the ability to deny tool calls. In practice, do not turn on auto-approve for write tools.
  • Beware of what the AI reads. Ploy's read permission includes form submissions and visitor data, which come from strangers. A malicious form message could contain instructions aimed at your AI. Do not combine read access to form messages with publish access in the same session if you can avoid it.
  • Money moves need your OK inside Ploy. The docs say that before an action in a connected service that can move money, such as payments, refunds, discounts, invoices or ad budget changes, Ploy shows a card in its own chat. Your AI app cannot approve it. An approval covers one action and expires after an hour.
  • Use a test workspace. Choose One workspace and connect to a copy first.
  • Know your undo. Ploy says published versions are snapshots you can restore under Versions, then publish. Databases keep a 30-day rolling history that only owners and admins can restore. Practice a restore once.
  • Revoke when done. Open Integrations, the MCP tab, then Manage connected clients, and select Revoke access.
  • Customer data. Anything the AI app reads is also seen by that app's provider, under its own terms. If your forms hold sensitive details, keep read access off. Check local privacy law, which this page cannot cover.

Google's page on AI features in Search says there are no extra requirements to appear in AI Overviews or AI Mode. So connecting an AI editor is not a ranking tactic. Judge it as a work tool.

If it does not work

  • The sign-in window does not open: allow pop-ups for your AI app and try again.
  • Ploy's tools do not show up: start a new chat or restart the app after connecting.
  • "List my workspaces" fails: you chose One workspace. Ask about that workspace directly, or reconnect with All my workspaces.
  • The app asks for an API key: Ploy needs none. Choose OAuth or the sign-in option.
  • Workspace API tokens are rejected: the MCP server does not use them.

Who should skip this

  • You do not use Ploy for your site. This connects to a Ploy workspace only. Other site builders, such as Webflow and Framer, list an MCP server on their own pricing pages, so check whether your platform has one.
  • Your plan blocks it. ChatGPT Pro is read-only, and mobile ChatGPT is not supported.
  • Nobody can review changes before they go live.
  • Your site has sensitive customer data in forms and you have no policy for AI tools reading it.

The fair view: it is often quicker to use Ploy's own chat for site edits. MCP is worth it if you already live in Claude, ChatGPT or Cursor and want one place to ask questions across tools, or if you want a coding agent to work on the site source and the site together.

How I'd approach this

Run a week-long trial with fixed rules and a written pass or fail.

  1. Day 1. Create a test workspace in a free Ploy account, with a copy of a simple page, not your live site.
  2. Day 2. Connect your AI app with One workspace and only Read and Chat. Ask for a list of sites and a screenshot. Pass: it answers correctly with no changes.
  3. Day 3. Ask for three small edits, such as a headline, a button label and a new FAQ. Review each in Ploy before you publish. Note the credits used.
  4. Day 4. Add Write only if you need table edits. Test on a throwaway table. Practice a restore.
  5. Day 5. Leave Publish off. Publish by hand after your review.
  6. Day 6. Try a bad prompt on purpose, such as "delete the pricing page", and confirm you can block and undo it.
  7. Day 7. Decide. Pass: edits match what you asked, nothing changed without your sign-off, undo works, credits are predictable. Fail on any one of these: keep MCP off and use Ploy's chat.

Only add Publish for a narrow, trusted case, and keep a person checking every release. If you want a team plan walk-through, you can book a Ploy demo and ask which permissions your roles should hold. If you need a custom AI agent that uses your business knowledge and connects to your CRM and calendar, Blur Studio builds that, and you can see what that includes. To get a workspace to test with, start with Ploy's free plan and run the Day 1 to Day 3 steps first.

For a closer look, see how to update client websites by chatting with Claude or ChatGPT.

For a closer look, see using Claude and ChatGPT with your website safely.

Questions people ask

What is the Ploy MCP server address?

As of November 2026, Ploy's docs give https://ploy.ai/api/mcp. It is a remote server (Streamable HTTP) that uses OAuth sign-in. You do not need an API key or token.

Does Ploy MCP work with ChatGPT?

Yes, on the web only. Full access, including changes to data, needs ChatGPT Business, Enterprise or Edu. On ChatGPT Pro, Ploy can only read data. It is not available in the ChatGPT mobile apps.

Does Ploy MCP work with Claude and Cursor?

Yes. In Claude Desktop or claude.ai you add a custom connector. In Claude Code you run one command. In Cursor you add the URL to the MCP configuration file and select Connect. A free Claude plan allows one custom connector.

Is it safe to let an AI app edit my website?

It can be, with limits. Start with read access, use a test workspace, keep publish off, and review every change. Vendors say custom MCP servers are not verified by them and that prompt injection is possible. See AI agent guardrails for small business. Check your own privacy duties under local law.

Can the AI app publish to my live site?

Only if you grant the Publish permission when you connect. You can leave it off and publish by hand in Ploy. To change permissions later, remove Ploy from the AI app and connect again.

How do I disconnect an AI app from Ploy?

In Ploy, open Integrations, choose the MCP tab, select Manage connected clients and choose Revoke access next to the app. You can also remove Ploy inside the AI app.

Does Ploy MCP cost extra?

Ploy's docs say it is included with every account. Work done through it uses the same AI credits as work in Ploy's own chat. See Ploy pricing explained.

What is the difference between Ploy MCP and WebMCP?

They are different. Ploy MCP lets an AI app work on your Ploy workspace. WebMCP, per Ploy's Ploybook library, is a separate setup that exposes selected website actions to AI assistants. Do not mix them up.

More in this topic: Ploy.ai: the product, pricing and reviews (12 guides) →

Sources

Ready for the next step?

A 30 minute call. If you do not need me, I will tell you.

Ranjit Bhinge You talk to me, Ranjit. I reply within a few hours.

Not sure what you need? Tell me what is going on. I will point you the right way, even if it is not me.

Blur Studio · 30 minute call

Book a 30 minute call

You talk to me, Ranjit. If you do not need me, I will tell you.

  • We talk about your business and what is going on.
  • You leave with a clear next step, even if it is not me.
  • I reply to every message within a few hours.
Loading the calendar…