AI Skill and MCP

Agents write good LAYR when they have three things: the rules of the language, exact references for every widget and diagnostic, and a way to check their work. LAYR ships all three.

#The LAYR Skill

The Skill is a folder of Markdown in the Agent Skills format: SKILL.md with the mental model, the canonical form and the mistakes agents make, plus references generated from the same schema as this site.

layr skills install

This copies the Skill into .claude/skills/layr/ and adds a short block to AGENTS.md that points other agents at it. layr skills update refreshes it after you upgrade LAYR, and layr skills path prints where it lives. The Skill's version always matches the installed @dynshift/layr.

#The workflow agents follow

1. Write the file.
2. layr format to rewrite it in canonical form.
3. layr analyze --json and fix every diagnostic, using layr explain <code> when a message is not enough.

The Skill teaches this loop, so an agent checks its own work the same way you would.

#The MCP server

layr mcp

layr mcp starts a Model Context Protocol server over stdio. Its tools are list_widgets, get_widget, explain_error, check_snippet, format, analyze_project and search_addons. Add it to any MCP client with the command layr mcp, run in your project.

#llms.txt

/llms.txt indexes these docs for agents that read the web, and /llms-full.txt is the whole documentation in one file. Every docs page is also available as Markdown at its URL plus .md.