{
  "id": "product-how-it-works",
  "title": "How Embassia Works — 4-Step Pipeline",
  "content_type": "feature",
  "canonical_url": "https://embassia.com/#how-it-works",
  "updated_at": "2026-09-28T13:00:00Z",
  "language": "en",
  "content": {
    "headline": "From HTML to Agent-Ready in Four Steps",
    "steps": [
      {
        "step": 1,
        "action": "Convert to endpoints",
        "description": "Turn your website pages into clean, typed API endpoints that machines can instantly read. Every page becomes a structured JSON document with stable IDs, canonical URLs, content hashes, and freshness metadata.",
        "technical_detail": "Embassia normalizes HTML content into typed JSON schemas. Each document includes: stable identifier, canonical URL, content version, last-modified timestamp, content hash, and a structured body."
      },
      {
        "step": 2,
        "action": "Deploy On-Site Agent",
        "description": "An on-site agent runs directly on your website, intercepts LLM crawlers, and serves structured data instead of raw HTML. The agent detects agent user-agents (GPTBot, ClaudeBot, PerplexityBot, etc.) and routes them to machine-readable responses.",
        "technical_detail": "The on-site agent sits at the edge layer (CDN/reverse proxy) and uses User-Agent detection to identify AI crawlers. Identified agents receive structured JSON responses; human browsers get the normal HTML page."
      },
      {
        "step": 3,
        "action": "Enable Agentic Protocol",
        "description": "Support Anthropic MCP, OpenAI function calling, and standard REST interfaces for maximum compatibility across all AI systems. Agents discover your site's capabilities through standardized manifests.",
        "technical_detail": "Embassia publishes discovery documents at standard well-known URLs: /.well-known/embassia.json, /.well-known/mcp.json, /llms.txt, /api/v1/openapi.json. These documents list all available tools, resources, and content endpoints."
      },
      {
        "step": 4,
        "action": "Capture Telemetry",
        "description": "Collect privacy-preserving analytics on agent interactions to understand what your market is actually asking. See which AI models visit, what questions they relay, and where your content falls short.",
        "technical_detail": "Telemetry is collected at the edge without exposing end-user PII. Aggregated insights include: crawler identity, query patterns, content hit/miss ratios, and intent classification."
      }
    ]
  },
  "metadata": {
    "content_version": "1.0.0",
    "content_hash": "how-it-works-v1-20260928"
  }
}