Getting Started

AI and machine-readable documentation

Edit page
Use raw Markdown, llms.txt, and the documentation MCP server with AI development tools.

Customer Portal publishes the same maintained documentation in formats intended for search, coding agents, and other AI tools. These endpoints supplement the rendered site; the Markdown files remain the source of truth.

Raw Markdown

Every maintained page has a raw representation under /raw:

Rendered: https://nuxt-customer-portal.com/architecture/layers
Markdown: https://nuxt-customer-portal.com/raw/architecture/layers.md

The Copy page menu can copy either the page content or its canonical Markdown URL.

llms.txt

Use the generated indexes when a tool can read URLs but does not support MCP:

  • /llms.txt provides the documentation map and descriptions.
  • /llms-full.txt provides the maintained documentation as one combined context.

Prefer the smaller index first, then load only relevant pages. Use the full file for cross-cutting architecture or contributor questions that genuinely need the entire documentation set.

MCP server

The documentation MCP server is available at:

https://nuxt-customer-portal.com/mcp

Its read-only find-customer-portal-docs tool searches maintained pages and returns their canonical rendered and Markdown URLs. An empty query lists the catalog.

Every response also identifies the immutable public Customer Portal commit against which the documentation was checked. An agent should keep that revision with any extracted code or behavioral claim instead of assuming the moving master branch still matches.

The server does not modify a Customer Portal deployment, access tenant data, or expose application administration tools. It only helps clients discover public documentation.

Source fidelity

Product behavior changes quickly. Documentation pull requests should verify public routes, environment variables, policies, and workflows against the current Customer Portal source. The revision shown on each rendered page and returned by MCP is the public source baseline for the current site. The site’s automated checks keep the MCP catalog aligned with maintained page titles, descriptions, and routes.