React Three Fiber logo

React Three Fiber

Community
notsointresting
react-three-fiber

Build 3D and WebGL experiences in React using react-three-fiber (R3F) — Canvas setup, useFrame/useThree hooks, JSX Three.js elements, geometry/materials/lighting, scroll-driven 3D storytelling, and the Drei/physics/postprocessing ecosystem. Use this skill whenever the user wants a 3D scene, product configurator, interactive portfolio, WebGL background, particle effect, 3D model viewer, scroll-linked 3D animation, or mentions Three.js, R3F, @react-three/fiber, @react-three/drei, WebGL, GLSL shaders, or "3D website" in a React/Next.js context — even if they just describe wanting something "immersive" or "interactive in 3D" without naming the library. Also consult this skill before deciding whether to reach for 3D at all — it includes a stack-selection and anti-pattern checklist for when plain Three.js, Spline, or a 2D solution is a better fit than R3F.

Overview

Publishernotsointresting
Repositoryfrontend-inspo-mcp
Skill namereact-three-fiber
Stars
5
Forks
0
Bundled files
3
LicenseMIT
Links
  • Markdown instructions

    A SKILL.md file the model loads on demand, so it only costs tokens when a request actually matches.

  • Works with any LLM

    AI skills are plain Markdown, not provider-specific code, so this works with GPT, Claude, Gemini, Grok, or a local model.

  • 3 bundled files

    Scripts, templates, and references the model can read while it works. Files are read-only and never executed.

  • Open source

    Published by notsointresting on GitHub. Read the source before you install it.

Installation

Install the React Three Fiber AI skill in TypingMind to use it with any LLM, or drop it into another agent that reads SKILL.md.

1

Install in TypingMind

TypingMind installs a skill straight from its GitHub folder — it reads SKILL.md, bundles the resource files, and stores the result locally.

  1. Open the app and go to Plugins → Skills.
  2. Choose "Install from GitHub".
  3. Paste the skill folder URL below and confirm.
    https://github.com/notsointresting/frontend-inspo-mcp/tree/main/src/sources/r3f-content
  4. Enable the skill in any chat where you want it available.
Plugins → Skills → Add skill → From GitHub URL, then paste the folder URL and press Continue.
2

Install in another agent

Any agent that reads the Agent Skills format can use this skill — copy the folder into that agent's skills directory.

Claude Code — .claude/skills
git clone --depth 1 https://github.com/notsointresting/frontend-inspo-mcp.git /tmp/frontend-inspo-mcp
mkdir -p .claude/skills
cp -r /tmp/frontend-inspo-mcp/src/sources/r3f-content .claude/skills/react-three-fiber
Restart Claude Code after copying so it picks up the new skill.

Use it in TypingMind

Enable React Three Fiber in any TypingMind chat and the model takes it from there. Its name and description sit in the system prompt, and the moment a request matches, the model loads the full instructions itself — you never invoke it by hand, and it costs no tokens until it is actually used.

The model loads React Three Fiber on its own as soon as a request matches it.

Works with any AI model

AI skills are plain Markdown instructions rather than provider-specific code, so React Three Fiber is not tied to the model it was written for. Install it once in TypingMind and use it with GPT-5, Claude, Gemini, Grok, DeepSeek, Mistral, Llama, or a local model you run yourself — all on your own API keys.

  • Loaded only when it is needed

    The system prompt carries just the name and description. The instructions are fetched on the first matching request, so an idle skill costs nothing.

  • Switch models mid-chat

    Because the skill is instructions rather than code, changing model does not break it — the next model reads the same SKILL.md.

Skill instructions

This is the SKILL.md content the model loads. Read it before installing — a skill is instructions your model will follow.

React Three Fiber (R3F)

react-three-fiber is a React renderer for Three.js: <mesh /> becomes new THREE.Mesh(). It has no overhead versus plain Three.js (components render outside of React's own tree) and every Three.js feature works here without exception — install a new Three.js version and it's usable immediately, no waiting on R3F to catch up.

bash
npm install three @types/three @react-three/fiber

@react-three/fiber@8 pairs with react@18; @react-three/fiber@9 pairs with react@19. Pick the matching major version.

Before writing any code: should this even be 3D?

Don't skip this. 3D is expensive — in bytes, battery, and dev time — and it's easy to reach for it because it's impressive rather than because it serves the page. Run through this before scaffolding a Canvas:

  1. Would a well-shot image, video, or CSS/SVG animation do the job? If the answer is "the 3D would just look cool," that's a yellow flag, not a green light. 3D earns its place when it demonstrates depth or spatial relationships that a flat medium can't — a product configurator someone rotates, a real architectural walkthrough, a data structure that's genuinely 3D.
  2. Pick the right tool for the job, not the flashiest one:
ToolBest forLearning curveControl
Plain Three.jsFine-grained control, non-React apps, minimal bundleSteepTotal
React Three FiberAnything already in React/Next.js; component-driven scenes, state-reactive 3DModerate (need both React + Three.js)Full — it's just Three.js in JSX
SplineDesign-first teams without engineers, fast prototypesLowLimited outside the visual editor
CSS 3D transforms / <model-viewer>Simple product rotations, single static modelsLowLow
  1. Common anti-patterns to warn the user about (or avoid yourself):
    • 3D for its own sake — floating abstract shapes with no relation to content. Slows the site, drains mobile batteries, doesn't move the metric that matters. Ask "would an image work?" before building it.
    • Desktop-only 3D — most traffic is mobile. A scene that isn't budgeted for phones will drop frames or crash on low-end devices. Test on real hardware, reduce geometry/texture quality on mobile, and offer a static fallback (poster image) below a device/perf threshold.
    • No loading state — 3D assets take time. Without a progress indicator or skeleton, users think the page is broken and bounce. Always wrap asset loading in <Suspense> with a visible fallback (see Loading section below).
    • Unbounded asset budgets — uncompressed GLTFs, 4K textures on a hero background, no LOD/instancing for repeated meshes. See references/architecture-decisions.md for a fuller performance checklist.

If 3D is justified, continue below. If you're unsure, ask the user what the 3D needs to accomplish (showcase a product, tell a scroll-driven story, visualize data, add ambient texture) — that answer determines which reference file below to lean on.

Quick start

tsx
import { Canvas, useFrame } from '@react-three/fiber'
import { useRef, useState } from 'react'

function Box(props) {
  const ref = useRef()
  const [hovered, hover] = useState(false)
  const [clicked, click] = useState(false)

  // Runs every rendered frame — this is the render loop
  useFrame((state, delta) => (ref.current.rotation.x += delta))

  return (
    <mesh
      {...props}
      ref={ref}
      scale={clicked ? 1.5 : 1}
      onClick={() => click(!clicked)}
      onPointerOver={() => hover(true)}
      onPointerOut={() => hover(false)}>
      <boxGeometry args={[1, 1, 1]} />
      <meshStandardMaterial color={hovered ? 'hotpink' : 'orange'} />
    </mesh>
  )
}

export default function App() {
  return (
    <Canvas camera={{ position: [0, 0, 5], fov: 75 }}>
      <ambientLight intensity={Math.PI / 2} />
      <spotLight position={[10, 10, 10]} angle={0.15} penumbra={1} decay={0} intensity={Math.PI} />
      <pointLight position={[-10, -10, -10]} decay={0} intensity={Math.PI} />
      <Box position={[-1.2, 0, 0]} />
      <Box position={[1.2, 0, 0]} />
    </Canvas>
  )
}

TypeScript: install @types/three and type props as ThreeElements['mesh']. React Native: import from @react-three/fiber/native and configure Metro's assetExts for glb/png/jpg if using useGLTF/useTexture.

Core concepts

<Canvas>

The root component — creates the WebGL context, scene, camera, and renderer. Everything 3D goes inside it; everything 2D (UI, layout) stays outside it in normal DOM.

tsx
<Canvas
  camera={{ position: [0, 5, 10], fov: 75, near: 0.1, far: 1000 }}
  shadows
  dpr={[1, 2]}          // clamp device pixel ratio for perf
  gl={{ antialias: true, alpha: true }}
>
  {/* scene contents */}
</Canvas>

useFrame(callback)

Subscribes a component to the render loop. Runs on every frame with (state, delta) — state is the full R3F state object (scene, camera, gl, clock, pointer, etc.), delta is seconds since the last frame. Mutate refs directly here rather than using React state, or every frame will trigger a re-render.

tsx
useFrame((state, delta) => {
  meshRef.current.rotation.y += delta
})

Pass a second argument to control execution order across multiple useFrame calls: useFrame(callback, priority) — higher runs later.

useThree()

Read-only (unless you set it) access to the R3F state object outside useFrame — camera, scene, gl, size, viewport, invalidate, etc. Subscribe selectively to avoid unnecessary re-renders:

tsx
const camera = useThree((state) => state.camera)   // only re-renders when camera changes

JSX = Three.js constructors

Any Three.js class is available lowercased as a JSX tag: <mesh>, <boxGeometry>, <meshStandardMaterial>, <group>, <ambientLight>. Constructor arguments go in the args array prop: <boxGeometry args={[width, height, depth]} />. Non-constructor properties are just props: <meshStandardMaterial color="hotpink" roughness={0.4} />. Use attach="material" (usually automatic) when nesting isn't inferable.

Events

Meshes accept pointer events directly: onClick, onPointerOver/onPointerOut, onPointerDown/onPointerUp/onPointerMove, onPointerMissed (fires on the canvas background when nothing was hit), onWheel, onContextMenu. Event objects carry stopPropagation(), intersections, point, object, and camera for precise hit-testing. For camera controls and drag/selection patterns, reach for @react-three/drei's OrbitControls, TransformControls, or raycasting helpers rather than hand-rolling them.

Refs

useRef() gives direct access to the underlying Three.js object (a THREE.Mesh, THREE.Group, etc.) for imperative mutation inside useFrame or event handlers — this is the normal, idiomatic way to animate in R3F, not useState + re-render.

Loading assets (always with Suspense)

tsx
import { Suspense } from 'react'
import { useGLTF, Environment } from '@react-three/drei'

function Model() {
  const { scene } = useGLTF('/model.glb')
  return <primitive object={scene} />
}

export default function Scene() {
  return (
    <Canvas>
      <Suspense fallback={<LoadingFallback />}>
        <Model />
        <Environment preset="city" />
      </Suspense>
    </Canvas>
  )
}

useGLTF.preload('/model.glb') warms the cache before the component mounts. Never ship a 3D scene without a visible loading state — see the anti-patterns list above.

Reference files

Read these as needed rather than loading everything up front:

  • references/geometry-and-scenes.md — built-in geometries, BufferGeometry, instancing for repeated meshes, materials, lighting types and shadows, and basic animation patterns (useFrame, spring physics, GLTF animation playback).
  • references/scroll-storytelling.md — scroll-driven 3D: linking scroll position to camera moves, object transforms, and section-based "acts," using @react-three/drei's ScrollControls, plus craft-level guidance (one engineered peak, avoid a cue that never fully reveals, don't let the page have "dead scroll" that changes nothing on screen).
  • references/architecture-decisions.md — the fuller stack-selection framework, a mobile/performance checklist (draw calls, texture/model compression, LOD, dpr clamping), and a declarative JSON-driven-scene pattern for cases where scenes need to be generated or configured dynamically rather than hand-authored as JSX.

Ecosystem

There's a large, well-maintained ecosystem around R3F — reach for these before building something from scratch:

PackageUse for
@react-three/dreiGrab-bag of essential helpers (loaders, controls, environments, text, shapes) — check here first for almost anything
@react-three/postprocessingBloom, depth of field, vignette, color grading, custom screen-space effects
@react-three/rapier3D physics — rigid bodies, colliders, joints, character controllers
@react-three/xrVR/AR controllers and events
@react-three/gltfjsxTurns a GLTF file into a typed JSX component
@react-three/uikitWebGL-rendered UI components inside the 3D scene
@react-three/csgConstructive solid geometry (boolean mesh ops)
@react-three/test-rendererUnit-testing R3F scenes in Node
react-springSpring-physics animation, pairs naturally with R3F
zustand / jotai / valtioState management commonly used alongside R3F scenes
levaInstant debug/tweak GUI panels
maathMath helper grab-bag (easing, random, matrices)

Common mistakes to catch in review

  • Animating via useState + re-render every frame instead of mutating a ref inside useFrame — kills performance at scale.
  • Missing <Suspense> around anything that loads an asset (useGLTF, useTexture, useLoader).
  • No mobile perf budget: full-res desktop textures/models shipped unconditionally.
  • Reaching for 3D when a video or CSS animation would serve the actual goal just as well (see the checklist above).
  • Building custom camera-drag/raycasting logic that @react-three/drei already provides.

Bundled files

The model reads these on demand while the skill is loaded. They are exposed as readable files and are never executed.

Frequently asked questions

What does the React Three Fiber AI skill do?

Build 3D and WebGL experiences in React using react-three-fiber (R3F) — Canvas setup, useFrame/useThree hooks, JSX Three.js elements, geometry/materials/lighting, scroll-driven 3D storytelling, and the Drei/physics/postprocessing ecosystem. Use this skill whenever the user wants a 3D scene, product configurator, interactive portfolio, WebGL background, particle effect, 3D model viewer, scroll-linked 3D animation, or mentions Three.js, R3F, @react-three/fiber, @react-three/drei, WebGL, GLSL shaders, or "3D website" in a React/Next.js context — even if they just describe wanting something "im...

Why use React Three Fiber on TypingMind?

Because you install it once and use it with any model. React Three Fiber is plain Markdown rather than provider-specific code, so the same skill runs on GPT-5, Claude, Gemini, Grok, or a local model — and you can switch model mid-chat without it breaking. TypingMind runs on your own API keys, so you pay providers directly instead of a per-seat subscription, and your skills and chats stay in your own storage.

How do I install React Three Fiber in TypingMind?

Open Plugins → Skills → Install from GitHub in TypingMind and paste https://github.com/notsointresting/frontend-inspo-mcp/tree/main/src/sources/r3f-content. TypingMind reads its SKILL.md and bundles its files and installs it as a skill you can enable per chat.

Which AI models can use React Three Fiber?

Any model you connect in TypingMind. AI skills are plain Markdown instructions rather than provider-specific code, so GPT, Claude, Gemini, Grok, and local models can all load this skill when a request matches it.

How many AI models can I use with React Three Fiber?

As many as you like. As long as a model supports skills, you can use React Three Fiber with it — GPT, Claude, Gemini, Grok, DeepSeek, Mistral, Llama and more — all on TypingMind with your own API keys.

Is the React Three Fiber AI skill free?

Yes. It is published on GitHub by notsointresting under the MIT license. You only pay your own AI provider for the tokens you use.

What are AI skills?

An AI skill is a reusable instruction bundle that teaches an AI model how to do one specific task. It follows the open Agent Skills format: a SKILL.md file with a name and description, plus any scripts, templates or reference files the model may need. The model reads the instructions only when your request matches the skill, so an installed skill costs nothing until it is used.

How are AI skills different from plugins or MCP servers?

A plugin or MCP server gives a model new tools to call — code that runs somewhere and returns a result. An AI skill gives the model knowledge and process instead: how to approach a task, which steps to follow, what good output looks like. Skills are plain Markdown, so they need no server, no API key and no runtime, and they work with any model.

View all

Set up your own AI workspace now

Get notified about new features and future giveaways by subscribing to our newsletter 👇