Agent builder canvas
Describe an agent in a sentence and watch it become a graph of steps, then test it on the canvas: a packet travels the edges, every node reports what it did, the failing step opens a suggested fix, and the rerun passes to the end.
Agent workspacesNo. 2102 variationsFree with an account
Playing on its own
Variation
- The idea
- Describe an agent, watch it built as a graph, test it on the canvas and fix the failing step
- Autoplay
- Build turns the written description into six nodes that fade in as their edges draw; a step is selected to show its settings; Test run sends a packet along the graph with results on each node; Match the PO fails and the inspector opens on a suggested fix with a diff; Apply fix reruns it to a pass, the approval step is skipped, a node is dragged into place, and the agent is published.
- When you click
- Type a description and press Build (Enter); drag nodes, drag the background to pan, arrow keys nudge the selected node and Delete removes it; click blocks in the palette to add steps; Test run (T); Apply fix or Dismiss; Publish once the test passes. On a phone the graph is a vertical flow, the palette a chip row and the inspector a bottom sheet.
- Look
- flat canvas, 10 px nodes, hairlines, fading nodes, drawn edges, travelling dot, sans with mono results type
- Variations
- Graphite, Blueprint
- Files
- index.html, style.css, script.js and prompt.md. No framework, no CDN, works offline.
The prompt and the files
Paste the prompt into Claude, ChatGPT, Lovable or any coding agent to rebuild it in your own product, or take the three files as they are.
Loading the files