BIG FROOTS
p5.js AI coding agent
How it works About Try BiGGi
For p5.js creative coders

Vibe code in p5.js.
Actually learn.
Build visuals and mini games.

BiGGi is a p5.js AI agent that starts small and grows with you. Prompt a piece, see it run, tweak it — no copy-paste, no walls of code.

Not a chat. BiGGi is a structured environment where your live canvas is the context. Every module is editable. Every generation is targeted — only what changed.
See how it works
Hi, I'm BiGGi
made in BiGGi
p5

Made for p5.js

BiGGi's module structure maps directly to p5's own architecture — setup, draw, events. It makes sense because p5 is already modular.

→

Start small, grow iteratively

Start with a bouncing ball. Add gravity. Add color. Add interaction. Each prompt is one idea — not an overwhelming feature spec.

⟳

Vibe code that teaches

The modular output and inline explanations aren't just for reading — the structure invites you to tinker, break things, and learn from it.

The BiGGi loop

Not a one-shot chat. A guided iteration cycle — every prompt builds on what's already on the canvas.

Prompt a piece Describe one small thing to add or change
→
Modules update Only the affected module is regenerated
→
Test on canvas Run it live before committing anything
→
Accept or reject edits Keep changes or send them back for revision
reject → back to modules
canvas state becomes the next prompt's context
BiGGi — Creating a responsive grid of circles
BiGGi demo
BiGGi — Creating a snake mini-game
BiGGi demo

Built-in context management

BiGGi's RAG system keeps your live canvas as the context — so every prompt is targeted, not a full rewrite.

What's in context

Chat stacks every message — context grows quadratically. BiGGi only sends the current code state — context stays constant.

Chat-based BiGGi Turn 1 msg 1 Turn 2 msg 1 msg 2 Turn 3 msg 1 msg 2 msg 3 Turn 4 msg 1 msg 2 msg 3 msg 4 grows each turn ↗ quadratic growth Turn 1 current code Turn 2 current code Turn 3 current code Turn 4 current code grows slowly ≈ quasilinear growth Context ≈ current modules — grows slowly, not exponentially

Token efficiency

Regenerating one module instead of the full sketch keeps output lean — faster, cheaper, more focused.

No more "give me the full code" for redundant regenerations: we know where to incorporate the modifications — and we'll show you, so it feels easier for you to get motivated to code by yourself.

Prompt Completion
BiGGi
Chat-based
5k 2.5k 1k 0 T1 T2 T3 T4 Total: 8,050 5k 2.5k 1k 0 T1 T2 T3 T4 T5 Total: 12,950

Single-task sample — BiGGi plateaus while chat-based grows each turn. The example compares two snake mini-games, one made with BiGGi's context management, the other with a naive context management.

vs. asking ChatGPT for p5.js code

The difference between getting code dumped on you and actually building something.

Generic AI chat
function setup() {
  createCanvas(600, 400);
  for(let i=0; i<100; i++){
    particles.push({x:random(w),
    y:random(h),vx:random(-1,1),...
  })
  }
}
function draw() {
  background(0);
  particles.forEach(p => {
    p.x += p.vx; p.y += p.vy;
    if(p.x<0||p.x>w) p.vx*=-1;
Full sketch in one block — where do you even start?
Want to change the color? Regenerate everything.
No canvas — paste it somewhere else to see it run.
Losing context after a few turns.
BiGGi
Drawing & DisplayRender shapes, colors, animation
Input HandlingMouse, keyboard, touch events
Collision DetectionBoundary checks & object hits
Figure MovementVelocity, gravity, trajectories
Each module editable independently — touch only what you need.
Live canvas in the same view — test before you commit.
Canvas state = context. BiGGi always knows what's on screen.
Invite yourself to modify — the structure makes it approachable.

Start your first p5 sketch

Free during beta. No setup, no boilerplate — just prompt and run.

Try BiGGi free