Your agent. Your canvas.
Enable your own agents to draw beautiful canvases.
Give the agent you already use the skills to turn ideas into diagrams, boards and wireframes. Every shape stays editable. Every layout can be measured and improved.
- Claude CodeAnthropic
- CodexOpenAI
- CursorEditor + agent
- OpenCodeYour choice of model
- Gemini CLIGoogle
- Kimi CodeMoonshot AI
- DeepSeekThrough OpenCode
One prompt to get started
Give your agent a canvas.
Copy this prompt into your agent. It finds the latest stable release, installs the drawing skills and CLI, then creates a small canvas to check the setup.
Use a terminal-capable agent with Git, Node.js and npm. The prompt checks the selected release’s requirements. Start a new agent session after setup.
Beauty you can reason about
Good layouts have measurable qualities.
Blankpaper scores crossings, label readability, spacing, flow and how well a drawing fits on screen. Your agent can use that feedback to revise a canvas, instead of stopping at the first attempt.
Fewer crossings and arrows that avoid unrelated shapes.
Labels with space to breathe, at a useful viewing size.
Connections that follow the story your diagram tells.
Our scoring draws on Mooney et al.’s study of graph-drawing metrics and Purchase’s research on aesthetics and understanding, with additional checks for labels and containers. These are practical measures of readability; the final visual judgment stays with you.
Draw locally, for free.
No Blankpaper account or API key is needed for local drawing and scoring. Open the result in Blankpaper or Excalidraw and keep editing. Optional cloud publishing needs Blankpaper Pro and a separate sign-in.
Stay current. Stay in control.
The same prompt handles setup and future updates. It resolves the latest stable release, pins its exact commit and preserves existing skills. Backups and validation make updates deliberate, with the previous runtime kept for rollback.