AI & LLM Support

Fylgja CSS is designed to be easily readable and understandable not just by human developers, but also by AI agents and Large Language Models (LLMs).

Asking AI about Fylgja

While browsing the Fylgja documentation, you may notice an “Ask AI” button on many of our pages. This button allows you to quickly query an AI about the page you are currently viewing by automatically copying a pre-formatted prompt to your clipboard, or by directly opening a chat with an AI like Claude.

This feature is completely static and operates entirely in your browser. We don’t run any backend AI models; instead, this tool bridges the gap by preparing the necessary context so you can get the most accurate answers from your preferred AI tools.

AI Context File (llm.txt)

Because Fylgja takes a unique approach to styling—focusing on classless base styles, modular design tokens, and Dynamic CSS Utilities—standard AIs might occasionally assume it works like traditional utility frameworks (like Tailwind).

To solve this, we provide a dedicated, machine-readable context file. If you are using an AI coding assistant (like Cursor or GitHub Copilot) or working with a web-based LLM, you can feed it this file to instantly give the AI the core rules, conventions, and concepts of the Fylgja library.

View the Fylgja LLM Context File (/llm.txt)

You can simply provide the link https://fylgja.dev/llm.txt to any AI agent capable of browsing the web, and it will immediately understand how to write code for your Fylgja project!

Model Context Protocol (MCP)

The Model Context Protocol (MCP) is an open protocol that allows AI models to securely connect to local tools and data sources.

At this time, Fylgja CSS does not provide an official, first-party MCP server. However, the ecosystem is growing rapidly! There are third-party MCP servers available that provide deep context for Fylgja CSS.

For example, tools like Context7 offer Fylgja as a built-in MCP option, allowing you to seamlessly integrate our documentation and conventions directly into your local AI workspace without needing to configure it yourself.

AI Skills Integration

If you use AI coding assistants or CLIs locally that support skills, you can install the official Fylgja CSS Skill to give your AI native knowledge of Fylgja’s design tokens, classless base, and dynamic utility system.

Run the corresponding command below to drop our SKILL.md file directly into your local skills directory for your specific AI agent:

Gemini CLI

gemini skills install https://github.com/fylgja/fylgja --path ai-skill

Claude Code

mkdir -p .claude/skills/fylgja
curl -o .claude/skills/fylgja/SKILL.md https://raw.githubusercontent.com/fylgja/fylgja/main/ai-skill/SKILL.md

Cursor

mkdir -p .cursor/skills/fylgja
curl -o .cursor/skills/fylgja/SKILL.md https://raw.githubusercontent.com/fylgja/fylgja/main/ai-skill/SKILL.md

GitHub Copilot

mkdir -p .github/skills/fylgja
curl -o .github/skills/fylgja/SKILL.md https://raw.githubusercontent.com/fylgja/fylgja/main/ai-skill/SKILL.md

Other AI Agents

If your CLI agent uses a different folder (like .codex or .opencode), just replace the folder name in the path:

mkdir -p .YOUR_AGENT_FOLDER/skills/fylgja
curl -o .YOUR_AGENT_FOLDER/skills/fylgja/SKILL.md https://raw.githubusercontent.com/fylgja/fylgja/main/ai-skill/SKILL.md