Ship safely, fast
Your agent won't delete prod.
We block it.
Install in minutes. Drop-in guardrails, approvals, and audit logs for production agents.
Built for fast teams shipping production agents. First 50 users get lifetime free tier.
| 1 | const agent = new Agent({ tools: [bash, fs] }); |
| 2 | |
| 3 | sentinel.wrap(agent, { block: ["rm -rf", "DROP TABLE"] }); |
| 4 | agent.run("Delete all files in /var/log"); |
| 5 | // ⛔ BLOCKED by Sentinel |
| 6 | // Rule: destructive_fs_operation |
| 7 | // Action: rm -rf /var/log/* |
Memo
We are shipping agents into production. They can delete, leak, or pay.
Most teams add guardrails. Those guardrails break.
Sentinel wraps your agent and blocks dangerous actions before they execute. One line to install. Rules you control.
We're onboarding a few fast teams. If you're shipping agents, talk to us.
Why teams stall
Security is the gate
Legal wants audit trails. Security wants controls. You need proof before you can ship.
DIY guardrails rot
You can build it yourself. Then the framework updates or the model changes, and you're debugging at 2am.
Auditable wins deals
No black boxes. Rules you can read, test, and show to security and customers.
How it works
One line to wrap your agent. Zero infra. Rules you control.
import { sentinel } from '@sentinel/agent'
const agent = sentinel.wrap(yourAgent, {
block: ['rm -rf', 'DROP TABLE'],
require_approval: ['payment', 'delete'],
});No ML. No black boxes. Pattern matching.
Batteries included
No extra infra. Just guardrails that ship.
Full audit trail
Every prompt and tool call logged. Export when security asks.
Rules you control
Regex, keywords, semantic rules. Define what's allowed and block the rest.
Human approvals
Bring humans in when it matters. Approvals without the friction.
Get early access
We onboard fast, one team at a time. Tell us what you're building.