Use XenoKit with AI coding agents

XenoKit is built command-line first, so an AI coding agent can work with it the way you do in Terminal. Two things tell an agent that XenoKit compiles a project.

xenokit agents

xenokit agents prints a short Markdown brief fitted to the project it runs in: edit sources, never compiled output; don’t run sass, esbuild or postcss directly; what each rule compiles to where, with which tool versions; whether a watcher is running; and which files have errors or warnings right now. --json gives the same as one object.

A XenoKit section in AGENTS.md and CLAUDE.md

Turn on App Settings›Projects›Keep AI instructions in every project. XenoKit then keeps a short, marked section in each project’s AGENTS.md or CLAUDE.md, which agents read automatically. A project with neither file gets an AGENTS.md.

• XenoKit only writes between <!-- xenokit:start --> and <!-- xenokit:end -->.
• Each project’s own switch (Project Settings › General) adds or removes the section for that project alone.
• xenokit agents --write and --remove do the same from Terminal.

Install the command-line tool first (App Settings › Command Line): agents run xenokit in Terminal, and XenoKit warns you until it’s on your PATH.

The commands an agent needs

$ xenokit status .  # files with errors or warnings now
$ xenokit log --failed --full  # a failed run's complete output
$ xenokit run --file <path>  # compile now (through the watcher when one is running)
$ xenokit file <path>  # how one file compiles
$ xenokit validate  # after editing .xenokit.yml

Every command takes --json. xenokit log --failed --full prints plain text — the command, tool versions and full output — ready to paste into an AI chat or a forum post.