Key Facts
- ✓ The Figma-use CLI provides approximately 100 commands for creating and modifying design elements in Figma.
- ✓ JSX importing in the tool is approximately 100x faster than traditional plugin API imports, significantly improving iteration speed.
- ✓ The render command connects to Figma's internal multiplayer protocol via Chrome DevTools for enhanced performance with large object groups.
- ✓ The tool is built using Bun and Citty for the CLI, with an Elysia WebSocket proxy for real-time communication.
- ✓ Figma-use addresses the limitation of the official Figma MCP server, which is primarily read-only.
- ✓ The developer is actively seeking feedback on CLI ergonomics, missing commands, and JSX syntax naturalness.
Quick Summary
A new command-line interface tool called Figma-use has been developed to bridge the gap between AI agents and design software. Created by developer Dan, this tool allows AI to actively design within Figma rather than just reading files.
The CLI provides approximately 100 commands that enable AI agents to create shapes, text, frames, and components, modify styles, and export assets. This represents a significant shift from existing solutions that are primarily read-only or require inefficient JSON schemas.
Core Capabilities
The Figma-use CLI offers comprehensive functionality for AI-driven design workflows. Its command set covers the entire design process, from initial creation to final asset export.
Key capabilities include:
- Creating and modifying basic shapes and text elements
- Building complex frames and component systems
- Adjusting styles and properties dynamically
- Exporting finished assets in various formats
One of the standout features is the JSX importing capability, which is approximately 100x faster than traditional plugin API imports. This speed improvement is crucial for AI agents that need to iterate quickly on design concepts.
"I wanted AI to actually design — create buttons, build layouts, generate entire component systems."
— Dan, Developer of Figma-use
Technical Architecture
The tool's architecture combines modern technologies for optimal performance. The CLI itself is built using Bun and Citty, while an Elysia WebSocket proxy handles real-time communication.
A notable technical innovation is the render command, which connects directly to Figma's internal multiplayer protocol via Chrome DevTools. This approach provides enhanced performance when dealing with large groups of objects, a common scenario in complex design files.
The entire system is designed to work with any LLM coding assistant, making it versatile and compatible with various AI development workflows.
Addressing Design Limitations
The motivation behind creating Figma-use stems from limitations in existing solutions. The official Figma MCP server is primarily read-only, restricting AI agents to viewing files rather than creating them.
As the developer explains, the goal was to enable AI to actually design:
I wanted AI to actually design — create buttons, build layouts, generate entire component systems.
Existing alternatives either lacked write capabilities or required verbose JSON schemas that consume excessive tokens, making them inefficient for AI operations. This tool provides a more direct and efficient approach to AI-driven design.
Availability & Feedback
The tool is available for installation via Bun using the command bun install -g @dannote/figma-use. A demonstration video is available, showcasing the tool's capabilities in a 45-second overview.
The developer is actively seeking feedback on several aspects of the tool:
- CLI ergonomics and user experience
- Missing commands or functionality
- Naturalness of the JSX syntax implementation
Community discussion and feedback are being gathered through a dedicated comments section, providing an opportunity for users to contribute to the tool's evolution.
Looking Ahead
Figma-use represents a significant step forward in enabling AI agents to participate in creative design workflows. By providing a comprehensive command set and optimized performance, it addresses key limitations of previous approaches.
The tool's compatibility with any LLM coding assistant and its efficient handling of design operations position it as a valuable addition to the AI design ecosystem. As feedback is incorporated and the tool evolves, it may further expand the possibilities for AI-assisted design.
