MURAQIB · Clearance Sandbox

Should your agent
do that?

Every consequential action an AI agent takes should be cleared before it executes. Describe one below and watch the verdict — or drop your agent file and see which controls would govern it. ثقةٌ مهندَسة · Trust, Engineered.

Sandbox — simulated verdict, runs in your browser

or connect your agent

Drop your agent file here — agent.json · main.py · tools.json · openapi.yaml — and get its governance passport: the actions it can take, the MURAQIB controls that would engage, and a ready-made integration snippet. Analyzed locally in your browser. Your file is never uploaded, stored, or transmitted.

What this is — and is not. This console is a browser-side demonstration running a small illustrative subset of the MURAQIB control catalog. Verdicts here are simulated, nothing is sealed, and no data leaves your machine. On the production rail, every verdict is enforced pre-execution, sealed into a tamper-evident evidence record, and countersigned with an offline key — which anyone can verify independently. Architecture and threat model: Trust & Security.

Should your agentdo that?

Should your agent
do that?