MotionRail — Agent Readiness Plan

Implement all isitagentready.com checks for motionrail.jujiplay.com.

1. Discoverability

Check	File	Notes
robots.txt	doc/public/robots.txt	RFC 9309, AI crawler entries, Content-Signal, sitemap
Sitemap	VitePress built-in	Enable sitemap plugin in config.mts
Link headers	doc/public/_headers	describedby, service-doc, api-catalog Link relations
DNS-AID	Cloudflare DNS	_mcp._agents.motionrail.jujiplay.com + _index._agents.motionrail.jujiplay.com HTTPS records

2. Content Accessibility

Check	Action	Files
Markdown for Agents	Pages Function middleware — intercept Accept: text/markdown, extract content, convert to markdown via turndown	functions/_middleware.ts (Cloudflare Pages native TypeScript middleware)
llms.txt	Already generated in buildEnd hook	.vitepress/dist/llms.txt
llms-full.txt	Add to buildEnd hook — concatenate all doc markdown	.vitepress/dist/llms-full.txt

3. Bot Access Control

Check	In
AI bot rules	robots.txt — GPTBot, Claude-Web, etc.
Content Signals	robots.txt — Content-Signal: ai-train=yes, search=yes, ai-input=yes

4. Protocol Discovery

Check	File	Notes
MCP Server Card	doc/public/.well-known/mcp/server-card.json	SEP-2127, remotes → /mcp
Agent Skills	doc/public/.well-known/agent-skills/index.json	v0.2.0, empty skills for now
API Catalog	doc/public/.well-known/api-catalog	RFC 9727 linkset
MCP Server	doc/functions/mcp.ts	3 tools: get_package_info (npm), get_build_status (GitHub Actions), search_docs (via llms-full.txt). Pure Workers fetch(), no SDK needed. Path: POST /mcp
WebMCP	Skip	Static docs site; MCP + Markdown for Agents covers agent needs
OAuth / A2A	Skip	Public docs site, no auth

5. Commerce

Skip all — OSS library, free.

Files to Create

doc/
├── agent-readiness.md                    (this file)
├── public/
│   ├── robots.txt
│   ├── _headers
│   └── .well-known/
│       ├── api-catalog
│       ├── agent-skills/
│       │   └── index.json
│       └── mcp/
│           └── server-card.json

functions/
├── _middleware.ts                    (Cloudflare Pages native middleware)
└── mcp.ts                           (MCP server)

Build Pipeline Changes

doc/package.json

{
  "scripts": {
    "build": "vitepress build"
  },
  "devDependencies": {
    "@types/turndown": "^5.0.6"
  },
  "dependencies": {
    "turndown": "^7.2.4"
  }
}

doc/.gitignore

Remove functions/*.js (functions moved to repo root as .ts files).

doc/.vitepress/config.mts

Add llms-full.txt generation in buildEnd hook — read each sidebar doc's markdown, concatenate into single file at path.join(outDir, "llms-full.txt").

MCP Server Implementation

Same pattern as vistaview. Changes needed:

• Package name: motionrail (not vistaview)
• Repo: juji/motionrail (not juji/vistaview)
• Docs origin: motionrail.jujiplay.com (not vistaview.jujiplay.com)
• Search: read from /llms-full.txt instead of /llms-full.txt

Cloudflare Pages Config

• Build command: pnpm build (from doc/ directory)
• Build output: .vitepress/dist/
• Root directory: doc/
• Functions: functions/ at repo root