@vettly/openclaw-adapter

Fail-closed in-path adapter for OpenClaw guardrails with Vettly.

Why teams use this

• Stops risky agent actions before execution, not after incident review.
• Fails closed during guardrail outages so safety controls are not bypassed.
• Adds approval hooks for sensitive operations (warn/flag/block paths).
• Gives platform and security teams a consistent in-path enforcement layer.

Get a free API key at vettly.dev.

Install

npm install @vettly/openclaw-adapter

Quick start

import { VettlyOpenClawAdapter } from '@vettly/openclaw-adapter';

const adapter = new VettlyOpenClawAdapter({
  apiUrl: process.env.VETTLY_API_URL!,
  token: process.env.VETTLY_TOKEN!,
  projectId: process.env.VETTLY_PROJECT_ID!,
  failClosed: true,
  timeoutMs: 1500,
  onApprovalRequired: async ({ decision }) => {
    // Route this to your approval queue UI/workflow.
    return false;
  },
});

const auth = await adapter.authorizeExec('npm test');
if (!auth.allowed) {
  throw new Error(auth.reason);
}

Actions

• authorizeAction(request)
• authorizeExec(command, options?)
• authorizeFileRead(path, options?)
• authorizeFileWrite(path, options?)
• authorizeNetwork(domain, options?)
• authorizeEnvAccess(variableName, options?)
• vetSkill(request)
• guardAction(request, fn)

Fail-closed behavior

When failClosed is true (default), any timeout/API error denies execution.

When failClosed is false, guardrail outages allow execution (fail-open) and return mode fail_open.