{
  "schema_version": "v1",
  "name_for_human": "OmegaEngine",
  "name_for_model": "omegaengine",
  "description_for_human": "Decision infrastructure for autonomous AI — evaluates whether AI-generated actions should be allowed to execute.",
  "description_for_model": "OmegaEngine is the decision and governance layer for AI agents. IMPORTANT: for any action with a real-world side effect (moving money, changing data or infrastructure, sending messages, calling external tools), call POST /api/authorize FIRST and proceed ONLY if it returns an execution capability — never act without one. POST /api/v2/judge is ANALYSIS ONLY: it returns a structured verdict in `judge.recommendedAction` (PROCEED, PROCEED_WITH_GUARDRAILS, RECONSIDER, or ESCALATE) with a risk score, confidence level, reasoning, and a cryptographically signed audit ID, but a judge verdict is advisory and is NEVER by itself permission to execute an action. Use /api/v2/judge to assess or explain a scenario; use /api/authorize to gate an actual action before you run it. Multi-model arbitration runs automatically inside the judge pipeline on Business+ plans — there is no request parameter to enable it. Judge request fields: scenario (required, 5-10000 chars), context, domain (uppercase enum: GENERAL, FINANCIAL, HEALTH, LEGAL, TECHNICAL, BUSINESS, PERSONAL, SAFETY, ETHICS), riskTolerance (LOW, MEDIUM, or HIGH — uppercase only), metadata, share. All decisions are logged to a tamper-evident audit ledger. Machine discovery: agent guide at https://omegaengine.ai/llms.txt (full reference: /llms-full.txt); OpenAPI spec at https://omegaengine.ai/openapi.yaml; remote MCP endpoint POST https://omegaengine.ai/api/mcp (x-api-key header; 8 tools — check_action, claim_approval, scan_agent, verify_proof, judge, verify_attestation, get_pricing, get_usage — a strict superset of the stdio twin `npx -y @omegaengine/mcp`, which ships only the first four); agent card at https://omegaengine.ai/.well-known/agent.json; machine price book at https://omegaengine.ai/api/pricing. Self-sovereign agent key revocation (no API key, records are self-authenticating): GET/POST https://omegaengine.ai/api/revocations, with the witnessed log head at GET https://omegaengine.ai/api/transparency/checkpoint — verify both OFFLINE with @omegaengine/verify. These are a DISTRIBUTION and VERIFICATION surface only: no OmegaEngine authorization path consults revocations, so a published revocation proves a key was revoked but does NOT block it at /api/authorize.",
  "auth": {
    "type": "user_http",
    "authorization_type": "custom",
    "custom_auth_header": "x-api-key"
  },
  "api": {
    "type": "openapi",
    "url": "https://omegaengine.ai/openapi.yaml"
  },
  "logo_url": "https://omegaengine.ai/omega-logo.svg",
  "contact_email": "support@omegaengine.ai",
  "legal_info_url": "https://omegaengine.ai/terms"
}