{
  "name": "SentioGate BYOK AI Gateway Agent",
  "version": "1.0.0",
  "description": "Zero-markup BYOK AI gateway and multi-LLM router on a global edge network. Routes OpenAI-compatible completions by task tier, lowest price, and measured latency with automatic 429/5xx circuit-breaker failover and automatic context compression.",
  "url": "https://sentiogate.app",
  "documentationUrl": "https://sentiogate.app/docs",
  "provider": {
    "organization": "SentioGate",
    "url": "https://sentiogate.app"
  },
  "supportedInterfaces": [
    {
      "url": "https://api.sentiogate.dev/v1",
      "transport": "HTTPS",
      "protocol": "openapi",
      "schemaUrl": "https://sentiogate.app/openapi.json"
    }
  ],
  "capabilities": {
    "streaming": true,
    "pushNotifications": false,
    "stateTransitionHistory": false
  },
  "defaultInputModes": [
    "text/plain",
    "application/json"
  ],
  "defaultOutputModes": [
    "text/plain",
    "application/json",
    "text/event-stream"
  ],
  "skills": [
    {
      "id": "route-llm-completion",
      "name": "Cost-First Multi-LLM Routing & 429 Failover",
      "description": "Route OpenAI-compatible chat completions across OpenAI, Anthropic, Gemini, Groq, DeepSeek, Mistral, xAI, and custom HTTPS vLLM endpoints with 0% token markup and automatic fallback to another model if one fails.",
      "tags": [
        "llm-router",
        "ai-gateway",
        "byok",
        "openai-compatible",
        "failover"
      ]
    },
    {
      "id": "live-llm-benchmarks",
      "name": "Live Provider Latency & Error Telemetry",
      "description": "Query real-time 24h rolling-average TTFT, p50 latency, and HTTP 429 error rates across LLM providers.",
      "tags": [
        "benchmarks",
        "telemetry",
        "ttft",
        "latency"
      ]
    },
    {
      "id": "context-compression",
      "name": "Multi-Turn Prompt & Context Compression",
      "description": "Clean duplicate messages and whitespace automatically and compact long sessions automatically before they hit the model's context limit.",
      "tags": [
        "compression",
        "token-optimization",
        "semantic-cache"
      ]
    }
  ]
}