Skip to main content

Janus

MCP server for infrastructure automation.

Overview​

Janus is the central infrastructure automation layer for Rhea. It exposes 100+ MCP tools for:

  • DNS management (Cloudflare)
  • Container deployment (Coolify)
  • Secret management (Infisical)
  • Repository management (GitHub)
  • Uptime monitoring (Uptime Kuma)
  • SSO configuration (Authentik)

URL​

MCP Connection​

{
  "mcpServers": {
    "janus": {
      "command": "uvx",
      "args": ["--from", "janus-mcp", "janus-mcp"],
      "env": {
        "JANUS_API_URL": "https://janus.meetrhea.com",
        "JANUS_API_KEY": "<your-api-key>"
      }
    }
  }
}

Key Tool Categories​

CategoryToolsDescription
DNSjanus_dns_*A record management for meetrhea.com
Coolifyjanus_coolify_*App deployment, env vars, logs
Secretsjanus_secrets_*Infisical secret management
GitHubjanus_github_*Repository management
Statusjanus_status_*Service registry
Uptimejanus_uptime_*Monitor management
Ticketsjanus_ticket_*Work tracking
Devlogsjanus_devlog_*Development logs
Probejanus_probe*Adversarial QA reviews
Hancockjanus_hancock_*Consent management

Probe​

Adversarial QA tool for AI agent work review. Intentionally harsh - finds gaps, testing theater, and false confidence.

janus_probe(work_summary, file_paths, tests_written, verification_steps)
janus_probe_tests(test_code, feature_description)
janus_probe_quick(claim, evidence)

Returns structured verdicts: ship, needs_work, or do_not_ship.

Hancock​

Consent management for AI agent actions. Creates requests requiring human approval.

janus_hancock_request(requestor, action, description)
janus_hancock_check(request_id)
janus_hancock_pending()

Use for high-stakes operations where AI should not proceed without human sign-off.