AI & agents
These docs are built for AI tools as much as for people. Three machine-friendly surfaces are published alongside the site:
llms.txt
/llms.txt— an index of every documentation page with titles and links, following the llms.txt convention./llms-full.txt— the entire documentation in a single markdown file, ideal for dropping into a model's context.
Markdown twins
Every page is also published as raw markdown at the same path with a .md suffix — for example /installation.md. Point an agent at any docs URL and swap .html (or no extension) for .md to get clean markdown.
MCP server
A Model Context Protocol endpoint is available at:
https://capacitor.nativescript.org/mcpIt speaks the MCP Streamable HTTP transport and exposes three tools:
| Tool | What it does |
|---|---|
search_docs | Full-text search across the documentation, returns matching pages with snippets |
get_page | Fetch any docs page as raw markdown |
get_sitemap | The llms.txt index, for orientation |
Connect from Claude Code
bash
claude mcp add nativescript-capacitor --transport http https://capacitor.nativescript.org/mcpConnect from other MCP clients
Use the Streamable HTTP transport with the URL above — the server is stateless (no sessions, no SSE), so any spec-compliant client works.