Comparing Implementation for AI: TinyMCE vs Froala

TinyMCE AI is a fully managed plugin. Adding it requires one plugin declaration, one JWT endpoint, and a config block. No AI backend to build or maintain. Froala AI Assist is a request layer: the plugin handles the editor UI, but you build, host, and maintain the AI backend it points at. This guide compares both by examining what it actually takes to ship AI in your RTE.

AI capabilities: What each editor includes

TinyMCE AI vs Froala AI Assist

Time to implement TinyMCE AI

Level of effort for implementation: Low to Medium.

TinyMCE AI is a fully managed premium plugin (Essential tier and above) that adds Chat, Quick Actions, and Review directly into the RTE. The managed backend handles model routing, and you control which models are exposed to users and whether they can switch between the available models. Try out the TinyMCE AI demo now. 

If your app needs TinyMCE AI on-prem, get early access by contacting the TinyMCE team.

Supported models, selectable in the configuration:

• Google Gemini
• OpenAI ChatGPT
• Anthropic Claude

Installation process overview

Setup has three parts:

1. Add tinymceai to plugins. The default Silver theme toolbar includes the AI buttons automatically; if you're using a custom toolbar string, add tinymceai-chat, tinymceai-quickactions, or tinymceai-review explicitly.
2. Add tinymceai_token_provider to the editor config. This must be a function that fetches a signed JWT from your backend and returns { token: string }.
3. Implement the JWT endpoint. The token requires five claims: aud (your TinyMCE API key), sub (user ID), exp, iat, and auth. Guides are available: 
   1. TinyMCE AI with JWT authentication (Node.js) Guide 
   2. TinyMCE AI with JWT authentication (PHP) Guide 

Optional config in the same tinymce.init() call includes:

• Custom quick actions
• Model restrictions
• Source document uploads
• Review types
• content_id scoping

Chat and Review render as sidebars, so set tinymceai_sidebar_type: 'floating' if your layout requires the panel to be draggable outside the editor.

For teams that need to call TinyMCE AI outside the plugin UI (background processing, server-side generation), a TinyMCE AI REST API is also available as an alternative integration path.

TinyMCE AI configuration code sample

tinymce.init({
  selector: "textarea",
  plugins: "tinymceai",
  toolbar: "tinymceai-chat tinymceai-quickactions tinymceai-review",
  content_id: "document-123",
  tinymceai_default_model: "gemini-2-5-flash",
  tinymceai_allow_model_selection: true,
  tinymceai_reviews: [
    "ai-reviews-proofread",
    "ai-reviews-improve-clarity",
    "ai-reviews-change-tone",
  ],
  tinymceai_quickactions_custom: [
    {
      title: "Explain like I am five",
      prompt: "Explain the following text in simple terms.",
      type: "chat",
    },
  ],
  tinymceai_token_provider: () => {
    return fetch("/api/token").then((r) => r.json());
  },
});

📖 Want to read more? Check out the TinyMCE AI documentation to learn more about advanced configuration.

Time to implement Froala’s AI Assist

Froala's AI Assist plugin, introduced in Froala 5.1, adds two AI-powered toolbar buttons to the editor. Unlike a managed service, AI Assist is LLM-agnostic by design, so it doesn't ship with a connected AI backend. Instead, it provides a flexible request layer that you point at whatever endpoint you control.

The plugin supports two integration patterns:

• Direct endpoint configuration: 
  • Point aiAssistEndpoint at your API.
  • Set headers via aiAssistHeaders.
  • Map your API's field names using aiAssistDataKeys.
  • Define an aiAssistResponseParserPath to extract the response from your backend's JSON structure.
• Custom request handler: Implement aiAssistRequest as an async function for full control: inject user tokens, handle errors, validate responses, and trigger analytics before or after content is inserted using the aiAssist.beforeInsert and aiAssist.afterInsert events.

To activate AI features, you must also set aiSupplementalTermsAccepted: true in your config to confirm acceptance of Froala's AI legal terms.

AI Assist basic installation

Adding AI Assist requires four steps:

1. Include the plugin JS file (ai_assist.min.js) and CSS file (ai_assist.min.css) in your project, either from your local Froala installation or via CDN.
2. Add aiAssist and/or aiShortCuts to your toolbarButtons array. aiAssist opens the AI chat popup (also triggerable with Ctrl+Shift+I / Cmd+Shift+I); aiShortCuts opens the tone and translation dropdown.
3. Set aiSupplementalTermsAccepted: true.
4. Configure your AI backend connection, either via aiAssistEndpoint with supporting options, or via a custom aiAssistRequest function.

You still need to build and host the AI backend that the configuration points at. The Froala AI Assist plugin handles the editor-side wiring, not the model or infrastructure. 

Custom tone options, translation targets, and prompt templates are all configurable via aiAssistToneOptions, aiAssistTranslateOptions, and aiAssistPromptTemplate, but each requires its own implementation work.

Level of effort for basic implementation: Extra Large.

Wrap up: Try it yourself

TinyMCE AI is production-ready from day one. Chat, Quick Actions, and Review are all configurable through the same init script you're already using. If you want to see how TinyMCE AI works in your application, start a free 14 day trial of TinyMCE today.

FAQs

Does Froala AI Assist include a backend AI service?

No. The plugin handles the editor-side UI. What it doesn't provide is any AI infrastructure behind it. You supply the endpoint, the model, the authentication, the error handling, and the hosting. Froala gives you the wiring harness; you build the engine. 

Which is faster to implement: TinyMCE AI or Froala AI Assist?

TinyMCE AI is significantly faster. For teams with a short runway or a single sprint to ship AI features, that difference is measured in weeks, not hours.

Can Froala AI Assist use Claude or Gemini?

Yes, technically, because Froala AI Assist is model-agnostic by design. You point aiAssistEndpoint at whatever backend you control, so if you've built an endpoint that routes to Anthropic Claude or Google Gemini, the plugin will work with it. 

What's the difference between Froala AI Assist and TinyMCE AI?

The difference is where the responsibility sits. TinyMCE AI is primarily a managed service. Froala AI Assist is a frontend plugin that gives your editor a structured way to talk to an AI backend you build, host, and maintain. 

Does TinyMCE AI require you to manage prompt engineering?

Not for the built-in features. Chat, Quick Actions, and Review all ship with default prompts that work out of the box. Where prompt configuration becomes relevant is when you define custom Quick Actions or custom Review types: those require a prompt string, but that's closer to filling in a config field than engineering a prompt chain. The managed backend handles model-specific formatting and context injection. For most teams, the prompt work is optional and additive, not a prerequisite to shipping.