Before your agent does something irreversible, ScopeGate asks: was this authorized? One API call. Signed permit or denial. Audit trail included.
Starts free. Metered billing. No lock-in.// Before every consequential agent action: const result = await sg.verify({ grantId: grant.grant_id, agentId: 'my-agent', requestedAction: 'send_email' }); if (!result.permitted) throw new Error('Denied'); // ✅ authorized — safe to proceed
How It Works
Your user delegates a task to the agent — specifying exactly which actions are permitted, and for how long.
Before every consequential action, the agent calls ScopeGate. One POST request. Returns permitted or denied with a signed token.
Every verification is logged — what was requested, who requested it, whether it was permitted. Nothing hidden, ever.
Why ScopeGate
Every permit is HMAC-signed. Agents can't forge authorization — and you can verify any decision independently.
npm install scopegate-client and you're verifying actions in minutes. No agent framework lock-in.
Every verification is timestamped and logged. When someone asks "did the agent have permission to do that?" — you have the answer.
Pricing
FAQ
/grant/verify counts as one verification. You're billed monthly based on total verifications. Starter is $0.001 per verify, Growth is $0.0005, Scale is $0.0002. A typical agent doing 10,000 verifications/month costs $10 on Starter or $5 on Growth.{ permitted: false, reason: "..." } with a specific reason — grant expired, action not in scope, agent mismatch, etc. Your agent handles the denial. Every denial is logged to the audit trail.