@runestone-labs/gatekeeper-client

Runestone Gatekeeper TypeScript Client

A minimal, framework-agnostic TypeScript client for Runestone Gatekeeper.

Installation

npm install @runestone-labs/gatekeeper-client

Usage

import { GatekeeperClient } from '@runestone-labs/gatekeeper-client';

// Initialize client
const gk = new GatekeeperClient({
  baseUrl: 'http://127.0.0.1:3847',
  agentName: 'my-agent',
  agentRole: 'openclaw',
  runId: 'optional-correlation-id'
});

// agentRole is required unless you set GATEKEEPER_ROLE in the environment.

// Or simply (requires GATEKEEPER_ROLE):
const gk = new GatekeeperClient('http://127.0.0.1:3847');

// Execute a shell command
const result = await gk.shellExec({ command: 'ls -la' });

if (result.decision === 'allow') {
  // Command was executed
  console.log(result.result.stdout);
} else if (result.decision === 'approve') {
  // Human approval required
  console.log('Waiting for approval:', result.approvalId);
  console.log('Expires:', result.expiresAt);
} else {
  // Denied by policy
  console.error('Denied:', result.humanExplanation);
}

Available Methods

shellExec(args)

Execute a shell command.

await gk.shellExec({
  command: 'npm install',
  cwd: '/path/to/project',
  timeoutMs: 30000
});

filesWrite(args)

Write a file.

await gk.filesWrite({
  path: '/tmp/output.txt',
  content: 'Hello, world!',
  encoding: 'utf8'
});

httpRequest(args)

Make an HTTP request.

await gk.httpRequest({
  url: 'https://api.example.com/data',
  method: 'POST',
  headers: { 'Authorization': 'Bearer token' },
  body: JSON.stringify({ key: 'value' })
});

callTool(tool, args)

Low-level method for calling any tool.

await gk.callTool('shell.exec', { command: 'echo hello' }, {
  idempotencyKey: 'req-123',
  origin: 'external_content',
  taint: ['external'],
  dryRun: true
});

health()

Check if Gatekeeper is running.

const status = await gk.health();
console.log(status.version, status.demoMode);

Response Types

All methods return a GatekeeperResult<T>:

interface GatekeeperResult<T> {
  decision: 'allow' | 'approve' | 'deny';
  requestId: string;
  reasonCode?: string;
  humanExplanation?: string;
  remediation?: string;
  policyVersion?: string;
  idempotencyKey?: string;

  // When decision === 'allow'
  result?: T;
  success?: boolean;
  executionReceipt?: ExecutionReceipt;

  // When decision === 'approve'
  approvalId?: string;
  expiresAt?: string;
  approvalRequest?: ApprovalRequestDetails;

  // When decision === 'deny'
  denial?: DenialDetails;
}

Integration Examples

See the integrations/ directory for examples of wrapping this client for specific agent frameworks:

• OpenClaw - OpenClaw skill integration

License

Apache-2.0