Using these docs from AI agents

These docs are published in forms that coding agents can read cheaply. Everything below is only reachable from the company VPN.

MCP server (recommended)

https://docx.payability.com/api/mcp is a read-only MCP server (Streamable HTTP, no auth beyond the VPN). It has four tools:

Tool What it does

list_components

Every documented system, with a description, page count and URL.

list_pages

The pages of one component in navigation order, with path, title and description.

get_page

The full AsciiDoc source of one page, by URL or by component + path.

search_docs

Ranked full-text search over all pages, optionally within one component.

Claude Code

claude mcp add --transport http --scope user pyb-docs https://docx.payability.com/api/mcp

Cursor

Add to ~/.cursor/mcp.json (all projects) or .cursor/mcp.json (one repo):

{
  "mcpServers": {
    "pyb-docs": { "url": "https://docx.payability.com/api/mcp" }
  }
}

Other clients

Any client that supports Streamable HTTP can use the URL directly. For stdio-only clients, bridge it with npx mcp-remote https://docx.payability.com/api/mcp.

Static files

If you can’t use MCP, these files are rebuilt with the site on every deploy:

  • llms.txt: every component and page, with one-line descriptions and URLs.

  • llms-full.txt: the AsciiDoc source of every page in one file (about 3 MB).

  • docs-index.json: the same content as JSON (components, pages, sources).

  • sitemap.xml: every page URL.

A line like this in a repo’s CLAUDE.md or Cursor rules helps agents find them: "Payability system docs: use the pyb-docs MCP server, or fetch https://docx.payability.com/llms.txt."

How it’s built

extensions/llms-txt.js in the pyb-docs repo writes the static files during the Antora build. api/mcp.mjs is a Vercel function that serves the MCP tools from the bundled docs-index.json. Page sources come back exactly as written, so attributes like {project-name} and include:: lines are not expanded.