Your agents need a dynamic sandbox.Read the blog post →

Policy demo
Block destructive changes with an intent-based policy

Agents can do serious damage, like wiping a production database or turning off backups. Instead of trying to list every risky command, you describe what you don’t want in plain English, and Qpoint checks each action against that description. In the demo, a request to reset a database without a backup is blocked in both Claude Code and Grok.
Gaurav Mahajan
Gaurav Mahajan
October 23, 2026

About this demo

Qpoint is runtime security for AI agents: a real-time, dynamic sandbox for every agent on a machine, including Claude Code, Codex, Cursor, Grok, or your own scripts. There’s no wrapper or proxy to set up and nothing to configure per agent.

In this demo, we write one intent-based policy in plain English: deny any action that destroys or overwrites production data or infrastructure, or disables the safeguards protecting it. We don’t list commands. Every action the agent takes is scored against that description by a fast decision model, and anything above the threshold is denied. We also show a static rule that blocks reads of SSH keys with no model involved.

Then we ask Claude Code to reset a database without a backup. The reset is blocked. We run the same prompt in Grok and get the same result: one policy, written once, enforced on two different agents.

See Qpoint in action in your environment
Request access and we'll get you set up.
Request Access