Labyrinth ScoutSecurity control plane for autonomous AI agents.

Prompt injection detection, kill switch, honeypots, ERC-8004/8126 on Base, and live monitoring. Self-host free or subscribe for hosted (£79/mo).

Prompt Guard

Three-layer injection detection on user input and tool output before it reaches your model.

Suspend any agent instance over Redis pub/sub. Carina stops processing within one message.

Radar view, event stream, anomaly timeline, and tool heatmap. Hosted subscribers receive a private console URL by email.

Node and Python clients for custom agents. Same API Carina uses natively.

Drop-in proxy for OpenClaw, Hermes, and other OpenAI-compatible stacks.

Need Scout pre-configured with Carina? Contact us for setup support.