@agentbouncr/core

The governance layer for AI agents. Deterministic rules — not another AI — decide what each agent can do. Policy Engine, Audit Trail, Kill-Switch, Event System, and Injection Detection. Works with LangChain, Vercel AI SDK, OpenAI, CrewAI, or any agent framework.

Installation

npm install @agentbouncr/core

Quick Start

import { GovernanceMiddleware } from '@agentbouncr/core';

const governance = new GovernanceMiddleware();

governance.setPolicy({
  name: 'basic-security',
  version: '1.0',
  rules: [
    {
      tool: 'file_write',
      effect: 'deny',
      condition: { path: { startsWith: '/etc/' } },
      reason: 'Writing to /etc/ is not permitted',
    },
    { tool: '*', effect: 'allow' },
  ],
  createdAt: new Date().toISOString(),
  updatedAt: new Date().toISOString(),
});

const result = await governance.evaluate({
  agentId: 'my-agent',
  tool: 'file_write',
  params: { path: '/etc/passwd' },
});

console.log(result.allowed); // false

For full documentation, examples, and architecture overview, see the main repository.

Upgrading to 0.2.0

Version 0.2.0 introduces a validated Agent State Machine with 7 states (was 4). Key breaking changes:

• 'running' status renamed to 'active'
• New states: 'idle', 'suspended', 'decommissioned'
• emergencyStop() and resetKillSwitch() are now async
• Invalid state transitions throw GovernanceError (code: INVALID_TRANSITION)

See CHANGELOG.md for the full migration guide.

License

Elastic License 2.0 (ELv2) — see LICENSE