Figma costs $15/month per editor and has been slowly adding AI features on top. OpenPencil, a new MIT-licensed project, takes the opposite approach: build the AI capabilities first and wrap a full vector editor around them.
The tool runs as a web app or Electron desktop app and covers standard vector design territory: infinite canvas, shape tools, Bezier pen, boolean operations (union, subtract, intersect), auto-layout, component systems with instances and overrides, and multi-page documents. The design system supports variables and tokens with multi-theme variants. So far, that's a competent Figma alternative. The interesting part is the AI layer.
Prompt-to-Canvas With Parallel Agents
OpenPencil uses an orchestrator that breaks complex page designs into spatial sub-tasks, then dispatches concurrent AI agents to generate different sections simultaneously. You describe what you want in text, and multiple agents work in parallel to produce it on the canvas with a streaming animation so you can watch the design materialize.
The tool supports Claude, GPT-4o, Gemini, GitHub Copilot, and Codex CLI, automatically adjusting prompt complexity based on the model's capabilities. It also includes an MCP server (Model Context Protocol, a standard for connecting AI models to external tools), meaning you can control the design tool from your terminal or integrate it into coding workflows.
Code Export to 8 Platforms
The other standout feature is code export. A single .op file (JSON-based and Git-friendly) can export to React with Tailwind, HTML/CSS, Vue, Svelte, Flutter, SwiftUI, Jetpack Compose, and React Native. That makes it genuinely useful for developers who want to design in a visual tool and ship real code, not just hand off mockups.
Figma import is already working, preserving layout, fills, strokes, effects, text, and vectors from .fig files. The project has 1,400 GitHub stars and 139 commits so far, with collaborative editing and a plugin system listed as upcoming.
OpenPencil is early, and the quality of AI-generated designs will depend heavily on which model you connect. But as a free, open-source design tool that treats AI as a core workflow rather than a bolted-on feature, it fills a gap that commercial tools are approaching from the other direction.