Agents

reelcn is built to be read by an agent as easily as by a person.

llms.txt

/llms.txt lists every item, grouped by category, with its one-line description, natural length, use/avoid notes and install URL — small enough to paste whole into a context window. /llms-full.txt adds a working usage example under each entry, generated from the same @example header as this site's item pages.

Every docs page is also available as raw markdown at its own URL with .md appended (/docs/theming.md), or by asking for it with Accept: text/markdown — useful for an agent that fetches a page and wants the source, not the rendered HTML.

shadcn MCP

The shadcn CLI ships an MCP server that lets an agent search and install registry items as tool calls instead of shell commands. Point it at reelcn by adding the registry to components.json first (see Installation):

{
  "registries": {
    "@reelcn": "https://www.reelcn.dev/r/{name}.json"
  }
}

Then wire the server into your agent:

npx shadcn mcp init

It asks which client to configure — Claude Code, Cursor, VS Code, Codex or OpenCode — and writes that client's MCP config for you.

Agent skill

The reelcn skill teaches an agent the whole job, not only installs:

• pick a video type
• choose a template or write a storyboard
• brand it
• review a contact sheet
• check determinism
• render

npx skills add Karanjot786/reelcn

It ships a catalog snapshot and eight recipes for:

• product launches
• feature shorts
• changelogs
• tutorials
• podcast clips
• talking-head shorts
• data stories
• YouTube intros

Each recipe gives a scene order, lengths, the items to use and a ready JSON example.

Next steps