Summary

Vibma empowers AI agents to design directly within Figma, enabling users to generate layouts, create components, modify styles, and build entire design systems through conversational prompts. It targets creators and designers seeking advanced AI-powered tools to significantly streamline the UI/UX development process, offering a novel paradigm for human-AI collaboration.

How It Works

The core of Vibma involves leveraging Large Language Models (LLMs) that are compatible with the MCP (Multi-modal Communication Protocol) to interpret natural language design requests and programmatically interact with the Figma API. This allows the system to read existing layouts, generate new, structurally sound components complete with instances and variants, dynamically modify styles, and construct comprehensive design systems. This conversational, API-driven approach aims to bridge the gap between abstract design intent expressed in natural language and functional, usable Figma assets.

Quick Start & Requirements

Two primary setup paths are available: cloning the repository and building from source (detailed in DRAGME.md), or a simpler installation directly via npm (detailed in CARRYME.md). Both methods necessitate an LLM that supports the MCP protocol. Model recommendations are task-dependent: GPT-5.3 Codex (medium reasoning) is suggested for cost-effective one-shot builds, while higher-tier LLMs such as GPT-5.3 Codex (xhigh), Gemini 3.1 Pro, or Claude Opus 4.6 are recommended for iterative design work where context persistence and multi-pass workflows are crucial.

Highlighted Details

• Demonstrates capability for achieving cost-effective one-shot design builds for under $1, utilizing specific LLMs like GPT-5.3 Codex (medium reasoning).
• Supports complex iterative design workflows with advanced LLMs such as Gemini 3.1 Pro and Claude Opus 4.6, ensuring consistent quality as conversational context grows.
• Prioritizes the generation of structurally usable Figma components, including proper instances and variants, distinguishing it from models that may only produce visually similar frames without underlying design system integrity.

Maintenance & Community

Vibma represents the inaugural open-source initiative from ufira, a platform dedicated to democratizing access to AI-powered creation tools for a broader audience. Community engagement, bug reporting, and feature requests are primarily managed through GitHub Issues, fostering collaborative development.

Licensing & Compatibility

The project is distributed under the permissive MIT License. This license permits broad usage, including commercial applications and integration into closed-source projects, without imposing significant restrictions.

Limitations & Caveats

A critical caveat is the necessity of selecting LLMs that reliably produce structurally sound Figma components; models that generate only superficial visual frames without proper instances, variants, or library support are explicitly discouraged and may result in unusable design assets. The overall effectiveness and output quality are intrinsically tied to the chosen LLM's capabilities and its adherence to the MCP protocol.