AI Support
Use Kikita UI with coding agents through curated Markdown context or the local read-only MCP server.
Use llms.txt for a concise index and llms-full.txt when an agent needs the full public documentation context.
Start with:
https://kikita-labs.github.io/kikita-ui-docs/llms.txt
Use full context only when the curated index is not enough:
https://kikita-labs.github.io/kikita-ui-docs/llms-full.txtRun the read-only stdio MCP server locally with npx. It exposes package docs, component examples, API sections, search, and prompt templates.
{
"mcpServers": {
"kikita-ui": {
"command": "npx",
"args": ["-y", "@kikita-labs/ui-mcp@latest"]
}
}
}Give agents a short rule that keeps them on published package APIs and the generated Kikita documentation surface.
Use Kikita UI docs through the kikita-ui MCP server.
Prefer package APIs and examples returned by the server.
Do not invent component inputs, outputs, CSS hooks, or imports.
If MCP is unavailable, use https://kikita-labs.github.io/kikita-ui-docs/llms.txt first, then llms-full.txt when full context is needed.The MCP package is generated from this docs site's installed @kikita-labs/ui version. If the package dependency changes, regenerate the agent surface before publishing docs or MCP data.
- Agents should import only from
@kikita-labs/ui. - Unreleased sibling repository source is not public documentation.
- The first MCP server is read-only: no file writes, shell commands, or telemetry.