Tiptap review: a headless editor framework you build the UI for

What Tiptap is, and who it is for

Tiptap is what you reach for when the editing surface has to be yours. It gives you ProseMirror's document model wrapped in an extension system that is far more pleasant than ProseMirror's own API, and then it stops: there is no toolbar, no bubble menu, no slash command, no file uploader. Those are components you write against commands the framework exposes.

That makes it a strong fit for products where the editor is part of a designed interface — a knowledge base, a CRM notes field, a CMS body editor, an AI writing surface — and a poor fit for a team that expected to install an editor and see one.

It is in this dataset because it is routinely shortlisted next to Plate, Editor.js and Lexical, and because those four are frequently confused with page builders. They are not: a requirement containing the words "columns", "breakpoints" or "brand colours" belongs with GrapesJS or Puck instead.

Architecture: extensions over ProseMirror

An extension can contribute nodes, marks, keyboard shortcuts, input rules, commands and rendering behaviour, and extensions compose into a schema. StarterKit is a bundle of the common ones — paragraphs, headings, lists, bold, italic, history — which is why the getting-started example is so short.

Underneath sits ProseMirror, and that matters in both directions. It gives you a battle-tested, transaction-based document model with strong guarantees about valid state. It also means that when something behaves unexpectedly, the concepts you need are ProseMirror's: schema, transactions, plugins, decorations. Tiptap does not hide that permanently, and the documentation stops short of where ProseMirror's begins.

The framework-agnostic core with first-party React and Vue bindings is worth noting. Of the headless document-editing frameworks in this dataset, only Tiptap and Editor.js can serve a Vue product without community glue. The finished editors, CKEditor 5 and TinyMCE, also ship first-party Vue integrations — with a toolbar you did not design and a GPL-or-commercial licence attached.

Getting started

This is the file from our benchmark, unedited. Eleven lines — the shortest React integration of the eight applications we measured.

import { createRoot } from 'react-dom/client';
import { EditorContent, useEditor } from '@tiptap/react';
import StarterKit from '@tiptap/starter-kit';

function App() {
  const editor = useEditor({
    extensions: [StarterKit],
    content: '<h1>Hello editor</h1><p>Type here.</p>',
  });
  return <EditorContent editor={editor} style={{ padding: 16 }} />;
}

createRoot(document.getElementById('app')).render(<App />);

Median 171 ms to a rendered editor. Compare that with Plate at 209 ms and Lexical at 159 ms: within this group the timings are close enough that they should not decide anything, and the benchmark page explains why we publish them anyway.

Strengths

The extension model is the whole API. Configure a list, get an editor. Adding a custom node type is one file, and the typed command API means mistakes surface at compile time.

Framework-agnostic core. First-party React and Vue bindings from one editing core, which no other headless rich-text framework here offers.

ProseMirror underneath. A document model with real guarantees, years of edge cases handled, and an escape hatch to raw ProseMirror plugins when the extension API is not enough.

Documentation organised by task. Each extension has a runnable example and an explicit install path — better than most libraries in this dataset.

A clear upgrade path for hard features. Collaboration, comments and AI exist as maintained products rather than as abandoned community experiments.

Limitations

No interface, at all. Toolbars, bubble menus, slash commands, link dialogs and upload handling are your code. For a designed product this is the point; for a team on a deadline it is a surprise, and it is the single most common reason Tiptap gets swapped out again.

Heavier than the minimal comparison suggests. 123.3 kB gzip for the React binding plus StarterKit, against Lexical at 104.9 kB and Editor.js at 64.3 kB. Realistic editors add extensions on top of all three.

ProseMirror leaks through. Not a flaw, but a cost: debugging unusual behaviour means learning a second mental model, and the documentation hands you off at that boundary.

Some expected capabilities are commercial. Collaboration, comments, AI and document conversion are paid services. That is a legitimate business model and it is worth establishing before a proof of concept assumes they are free.

Not a page builder. No canvas, no layout, no styling for end users. Worth repeating because the category names overlap.

Pricing

The framework and its open-source extensions are MIT licensed, free, with no ceiling. The hosted services are separately priced; our verification pass did not capture the current figures, so the dataset records paid_from_usd as null rather than an estimate.

For a hands-on comparison against the closest alternatives, see Tiptap vs Lexical, Plate vs Tiptap and Editor.js vs Tiptap.