AgentForge

Trust Model

Control which agents can communicate — allowed routes, confidence thresholds, and validation.

TrustConfig

interface TrustConfig {
  allowedHandoffs?: Record<string, string[]>;
  allowedRoutes?: Record<string, RouteConfig>;
  minConfidence?: number;
  validateOutputs?: boolean;
  requireExplicitRoutes?: boolean;
  fallbackAgent?: string;
  maxConfidenceVariance?: number;
  routeOverrides?: Record<string, RouteConfig>;
}

Allowed Handoffs

Restrict which agents can hand off to which:

trustModel: {
  allowedHandoffs: {
    triage: ['support', 'billing', 'technical'],
    support: ['escalation'],
    billing: ['escalation'],
    // technical has no allowed handoffs — it's a leaf
  },
  requireExplicitRoutes: true,
},

Minimum Confidence

Require a minimum confidence score for handoffs:

trustModel: {
  minConfidence: 0.7,  // Reject handoffs with < 70% confidence
},

RouteConfig

Per-route configuration:

interface RouteConfig {
  minConfidence?: number;
  maxHandoffs?: number;
  timeout?: number;
}
trustModel: {
  allowedRoutes: {
    'triage→support': { minConfidence: 0.8, timeout: 30_000 },
    'triage→billing': { minConfidence: 0.6 },
  },
},

Next Steps