The First Rich Text Editor With Native MCP Support (Cursor + Claude + ChatGPT)

Eddyter is the first rich text editor with a native MCP server. Connect it to Cursor, Claude or ChatGPT and your AI assistant can add the editor to your app, check your license, read your AI credit balance and switch editor features on or off, straight from chat. No dashboard tab. No copy-pasting API docs into a prompt.

That's the short version. The rest of this post covers what "native" means here (it matters, because a couple of editors now ship documentation MCP servers), what the Eddyter MCP server can do, and how to set it up in each client in about two minutes.

What is MCP? (The 40-word answer)

MCP (Model Context Protocol) is an open standard that lets AI assistants call external tools. An MCP server exposes a set of actions, such as "check license" or "update config", and clients like Cursor, Claude and ChatGPT can run those actions on your behalf from a normal chat prompt.

Think of it as a USB port for AI agents. The agent doesn't need to know how Eddyter's backend works. It just sees a list of tools with names and descriptions, and picks the right one.

Docs MCP vs. native MCP: why "first" is a fair claim

Here's something we noticed while researching this post. Search "rich text editor MCP" and you'll find a few results. TinyMCE now hosts a documentation MCP server. DHTMLX RichText has one too. There's also a community-built Tiptap MCP package on npm.

Those are useful. But look at what they actually do.

TinyMCE's own docs say its server "serves this documentation and nothing else." DHTMLX describes theirs as giving assistants "verified, current documentation." In both cases the agent gets smarter about the editor's API. It still can't touch your editor.

So we split it like this:

A native MCP server is one where the editor vendor exposes the product itself, your account and your live editor config, to the agent. By that definition, Eddyter is the first rich text editor we've found that ships one. If you know of another, tell us and we'll update this post. (We'd rather be corrected than wrong in public.)

What the Eddyter MCP server can do

The server exposes five tools. Four read, one writes.

The one that saves the most time is get_integration_snippet.

You know the bug. You drop a rich text editor into a Next.js App Router page, run the dev server, and get window is not defined. Ten minutes later you're on Stack Overflow reading about dynamic imports. When we asked the Eddyter MCP for a Next.js App Router snippet, it came back with two files: an Editor.tsx marked 'use client', and an EditorClient.tsx that loads it through next/dynamic with ssr: false. Plus the env variable name and a note on why SSR has to be off. That's the bug fixed before you ever hit it.

And update_config is the one people don't expect. Ask "turn off AI chat in the editor" and it's off in your live app. No redeploy. That works because Eddyter's feature toggles already live on the server (it's the same mechanism behind the runtime settings panel), so the MCP tool flips a switch that already existed.

On our own AI Pro test key, get_config lists over 50 toggleable features: tables, comments, @mentions, charts, PDF and DOC export, speech-to-text, version history, equations, email templates and more. Asking an agent "what's enabled?" beats clicking through a settings page.

How to set up the Eddyter MCP in Cursor, Claude and ChatGPT

Before any of these: create a free Eddyter account and make sure you have an active license key. The MCP server can't create keys for you.

The hosted server lives at https://mcp.eddyter.com/mcp and signs you in with OAuth, so for most clients you never paste a key at all.

Cursor

The fastest route is the one-click install from the Eddyter listing on cursor.directory. Sign in when prompted, pick your license key, done.

Prefer config files? Add this to .cursor/mcp.json:

1
2
3
4
5
6
7
8
{
  "mcpServers": {
    "eddyter": {
      "type": "http",
      "url": "https://mcp.eddyter.com/mcp"
    }
  }
}

Claude Code

Install it as a plugin:

1
2
/plugin marketplace add EddyterAI/eddyter-mcp
/plugin install eddyter@eddyter

Or run it locally with your key:

1
claude mcp add eddyter --env EDDYTER_LICENSE_KEY=eddyt_your_key_here -- npx -y eddyter-mcp

Claude.ai and Claude Desktop

1. Open Settings → Connectors → Add custom connector.
2. Paste https://mcp.eddyter.com/mcp.
3. Sign in with your Eddyter account and approve the license key.
4. Test it: "Check my Eddyter license status."

ChatGPT

ChatGPT supports custom MCP servers through Developer Mode.

1. Go to Settings → Apps & Connectors → Advanced settings and turn on Developer Mode. (On Business, Enterprise or Edu, a workspace admin may need to allow it first.)
2. Back in Apps & Connectors, click Create.
3. Name it "Eddyter", paste https://mcp.eddyter.com/mcp, and choose OAuth.
4. Sign in, let ChatGPT scan the tools, then click Create.
5. In a new chat, click +, pick Eddyter, and ask away.

One catch: OpenAI currently limits write actions in Developer Mode to some plans, so on certain accounts update_config may not be available. The four read tools work fine.

Codex (bonus)

1
codex mcp add eddyter --url https://mcp.eddyter.com/mcp

The full source and every config option are on GitHub (EddyterAI/eddyter-mcp). The server is MIT licensed.

Prompts you can paste right now

Once connected, these all work as plain chat:

• "Add Eddyter to my Next.js app."
• "Give me the SSR-safe setup for Vite."
• "Is my Eddyter key valid? When does my subscription renew?"
• "How many AI credits are left this month?"
• "What editor features do I have enabled?"
• "Turn off AI chat and turn on version history."

The first one is where it clicks. In Cursor, the agent calls get_integration_snippet, creates the component files in your project, adds the env variable, and tells you where to render it. Our homepage says setup takes 10 minutes. With the MCP, it's closer to two.

Safety: what the agent can and can't touch

Giving an AI agent write access to production config should make you a little nervous. Good. Here's how the server handles it.

• Local installs are read-only by default. update_config only appears if you set EDDYTER_ALLOW_WRITES=true.
• Plan limits are enforced server-side. The agent can't unlock a feature your plan doesn't include.
• It can't create or delete keys. Billing and key management stay in your dashboard.
• OAuth, not pasted secrets, for the hosted setup in Cursor, Claude, ChatGPT and Codex.

If you're on a team, our advice is simple: give the hosted connector to the people who own the editor config, and give everyone else a local install with writes off.

Who this is for (and who it isn't)

It's a good fit if you already write code with Cursor, Claude Code or Codex, you ship on React or Next.js, and you'd rather ask "what's my AI usage?" than open another dashboard.

It's less useful if you never use an AI coding assistant. Eddyter works the same without the MCP; the 10-minute manual setup in our quick install guide is still there.

Curious how Eddyter compares on everything else? Our top 10 rich text editors for developers in 2026 puts it next to Tiptap, CKEditor and TinyMCE. And if you're deciding how to save editor output, read HTML vs JSON vs Markdown for rich text content.

FAQ: rich text editor MCP

Does Eddyter work with Cursor?

Yes. Install it from cursor.directory with one click, or add https://mcp.eddyter.com/mcp to .cursor/mcp.json. Cursor signs you in with OAuth, so you don't paste a license key.

Can I use the Eddyter MCP server with ChatGPT?

Yes, through ChatGPT's Developer Mode. Create a custom app with the URL https://mcp.eddyter.com/mcp and OAuth authentication. Write actions depend on your ChatGPT plan.

Is the Eddyter MCP server free?

The MCP server is MIT licensed and free to use. You need an Eddyter account with an active license key, and the free plan works. See pricing for plan features.

What's the difference between a docs MCP and a native MCP?

A docs MCP lets an agent search an editor's documentation. A native MCP lets the agent act on your actual editor: check your license, read usage and change live feature settings.

Can the AI agent break my production editor?

It can only toggle features your plan allows, and changes apply to editor features, not your app code. Local installs have writes turned off unless you set EDDYTER_ALLOW_WRITES=true.

Which frameworks does the MCP generate setup code for?

React, Next.js (App Router and Pages Router), Vite and Remix. Eddyter itself also supports Vue, Angular, Svelte, Laravel, Vanilla JS and React Native.

Try it in two minutes

Get your free Eddyter API key, connect the MCP in whichever assistant you already use, and type "Add Eddyter to my app." Full setup docs are on the Eddyter MCP page.