Now available for custom environments

Write real code
without typing
a single word

Training Wheels is a visual programming shell that compiles to real code. Click shapes, build logic — zero keyboard required. Deploy it in any environment, for any audience.

See Plans & Pricing Try the Demo →
Training Wheels — interactive preview
+
DATA
ACTION
LOOP
LOGIC
Compiled Output
# training wheels
# user typed 0 characters
value_1 = []
# type: list
for item in value_1:
...
print(value_1)
Code Mirror
0×
Keystrokes required
100%
Real compiled output
Custom environments
1IR
Single source of truth

How It Works

Click shapes.
Build logic. Ship code.

Four steps from blank canvas to running program — no keyboard, no syntax errors, no guesswork.

STEP 01
Click +
The plus opens the L1 category wheel — four semantic shape families that map to the four pillars of any program.
STEP 02
Pick a category
Square = Data. Triangle = Action. Circle = Loop. Diamond = Logic. Every shape maps 1:1 to a real code concept.
STEP 03
Drill down
The L2 wheel appears with the specific options for that category. Pick one and it locks into your program.
STEP 04
Compile
The IR panel instantly emits Python (or whatever target language you configure). Zero syntax errors — structurally impossible.

Environments

One engine.
Any audience.

The shape library and compiler target are fully configurable. Swap the L1/L2 vocabulary and the emitter to fit your domain in minutes.

🎓
K–12 Education
Teach computational thinking before syntax. Students build real programs by choosing shapes, not memorising keywords. Works on tablets — no keyboard needed.
Education
Accessibility Tools
Motor accessibility, dyslexia support, or switch-access devices. Training Wheels is the first visual programming shell with zero typing surfaces by design.
Accessibility
No-Code Automation
Let non-technical teams build workflows without writing a line of code. White-label it inside your SaaS product with a custom shape vocabulary and your own emitter.
Automation
🧪
Rapid Prototyping
Sprint through logic scaffolding without touching a keyboard. Click your structure, export the IR, wire in the details later. Perfect for early ideation sessions.
Prototyping
🏭
Enterprise Workflow Builder
Map your business rules to a custom shape library. Compile to your internal DSL. Training Wheels gives non-developers a structured way to express logic without breaking production.
Enterprise
🎮
Game & Simulation Scripting
Let players or designers script game behaviors visually. Map shapes to game events, compile to your scripting engine. Runs on controller input — no keyboard on the couch.
Gaming

Features

Everything you need.
Nothing you don't.

Minimal surface, maximum power. The reference is a single HTML file — the React build adds structure for embedding.

🔇
Zero typing surfaces
No <input>, no <textarea>, no contenteditable. Click-only by architectural constraint.
🔷
Semantic shape vocabulary
Every shape maps 1:1 to a code concept. Squares are data. Triangles are actions. No decorative elements.
📋
IR as source of truth
The program array drives all panels. Compile output, code mirror, and strip all derive from the same IR — never from the DOM.
⚙️
Swappable compiler target
The emitter is a pure function over the IR. Swap it to target Python, JavaScript, Rust, your own DSL, or a JSON API call.
🎨
Custom shape library
Edit the L1/L2 arrays to add, remove, or rename any shape. No re-architecture needed — it's all data.
📦
Single-file reference
The canonical MVP is one HTML file with zero dependencies. Embed it anywhere. The React build is available for structured deployments.
🌀
Spring physics everywhere
One easing curve throughout: cubic-bezier(0.34, 1.56, 0.64, 1). Consistent, satisfying motion on every interaction.
🪞
Code Mirror panel
See the shape of your code as colored rectangles — loops and logic show indented bodies. Instant visual feedback on program structure.

Early Feedback

What people are saying

From educators, accessibility designers, and developers who've seen the reference.

This is the first visual tool I've seen where the shapes actually mean something. My students stopped asking why loops look like loops — they just click the circle.
🎓
CS Teacher, Middle School
Using the education environment
We white-labelled it in two days. Swapped the shape library to match our DSL, plugged in our emitter, and shipped. Non-technical stakeholders are now writing real workflow rules.
CTO, Workflow SaaS
Enterprise white-label
The zero-typing constraint isn't a limitation — it's the whole point. Users on switch access can build full programs now. That's never been possible before.
Accessibility Product Lead
Assistive technology integration

Pricing

Start simple.
Scale when you're ready.

Every plan includes access to the full engine. Higher tiers unlock white-labelling, API access, and custom compiler targets.

Solo
$29 / mo

For individuals and small projects exploring what Training Wheels can do.


  • Full reference implementation (HTML)
  • 1 custom environment config
  • Up to 3 users
  • Python emitter included
  • Email support
  • Custom compiler target
  • White-label / remove branding
  • API access
Get Solo
Enterprise
Custom

For organisations that need white-label, API access, and dedicated support.


  • Everything in Studio
  • Unlimited users
  • Full white-label (no Training Wheels branding)
  • REST / WebSocket API access to IR
  • Custom domain + CDN hosting option
  • SLA + dedicated Slack channel
  • Bespoke compiler integrations
  • Source code licence available
Contact Sales

All plans billed monthly. Annual billing available at 20% discount. Cancel any time.


FAQ

Common questions

Can't find your answer? Email us.

What is a "custom environment"?
An environment is a configured instance of Training Wheels with its own shape vocabulary (L1/L2 arrays) and compiler target. For example: an "automation" environment with shapes for triggers, conditions, and actions that emits JSON; or a "game scripting" environment that compiles to Lua. You edit a single config object — no core changes required.
Can it compile to languages other than Python?
Yes. The emitter is a pure function that receives the program IR array and returns a string. You can target Python, JavaScript, TypeScript, Rust, SQL, your own DSL, or a JSON payload for an API. Studio and Enterprise plans include support for building custom emitters.
Can I embed Training Wheels inside my own product?
Yes. The single-file HTML version drops into any page as an iframe or inline. The React build (Vite + Zustand) can be embedded as a component in any React app. Enterprise plans include a full source code licence for deeper integration.
What does "zero typing surfaces" really mean?
There is no <input>, <textarea>, or contenteditable element anywhere in the application — by architectural constraint, not convention. Every piece of program state comes from shape selections. This makes Training Wheels usable with switch access, eye-tracking, touchscreens, game controllers, and any other pointing device.
Is there a free trial?
The reference implementation (training-wheels-mvp.html) is freely available as a full working demo — no account needed. Paid plans unlock configuration, multi-environment deployment, and compiler customisation.
How do I get started with a custom shape library?
Edit the L1 and L2 data objects — each entry is just { id, shape, color, label }. The four shape types (sq/tri/cir/dia) remain the visual primitives; the labels and colors are yours to define. Studio onboarding includes a 1:1 walkthrough of this process.

Ready?

Start building without typing.

Try the free demo, or pick a plan and ship your own environment today.

View Plans Open Free Demo →