New: Decision Runtime — closed-set judgments for Agent OS. Read the guide →
HazelJS LogoHazelJS

AI Agent Governance

AI agent governance is how organizations control what agents may intend, decide, and do. HazelJS chains Skillgate (capabilities), Decision Runtime (closed-set choices), Gatekeeper (authorization), and timelines (audit).

Get Started with HazelJSView on GitHub

01Control path

Agent
 ↓
Intent
 ↓
Policy / Decision
 ↓
Authorization (Gatekeeper)
 ↓
Tool
 ↓
Audit (timelines)

02Components that exist today

  • Skillgate — curated skills from OpenAPI/modules; read/write/approval classes
  • Decision Runtime — closed-set decide(); confidence informs policy, never grants permission
  • Gatekeeper — allow / deny / require-approval before tool execution
  • DNA policies — packaged with the agent contract
  • Timelines / logs — JSONL run history via hazel logs

03Skillgate registration

typescript
import { Skillgate } from '@hazeljs/skillgate';
import { ToolRegistry } from '@hazeljs/agent';

const gate = Skillgate.fromOpenApi(spec, {
  // classify read / write / approval
});
gate.register(new ToolRegistry(), 'api-concierge');