Choose an Attack Scenario
Select a scenario to begin.
Two-Line Integration
The demo ends in the same place the product does: install Core and wrap the guarded outbound tool.
npm install @cerberus-ai/core
import { guard } from '@cerberus-ai/core';
const { executors: tools } = guard(rawTools, config, ['sendEmail']);
⚡
Start with 1. Run Control, then 2. Run Protected Attack
Watch the same runtime path move from normal behavior to a blocked outbound attack, then install Core from the docs.
What Cerberus Is Watching
Private Data Access
Reading sensitive files or records
Injection Detected
Hidden instructions in external content
Outbound Attempt
Trying to send data outside the org
Memory Poisoning
Malicious content stored for later
Threat Level
0/4
Cerberus Decision
WAITING
Incident Intelligence
Metrics Dashboard ↗
Select a scenario to see what's at stake.
Demo Progress
Start with 1. Run Control.
1
1. Run Control
Show a normal trusted-data workflow that completes without a block.
2
2. Run Protected Attack
Replay the lethal runtime path and watch Cerberus interrupt the outbound action.