Set one config value. Scout sees every tool call, every model decision, every workflow, and is ready to halt anything that goes wrong. Pre-wired for Carina. Connects to any agent via API.
One workflow from first draft to production. Carina is where the work happens; Scout is what keeps the same system safe when real work starts moving.
Open the workspace, work fast, and keep everything in one place.
Protect prompts, tools, and external calls when the work becomes sensitive.
Audit, policy control, egress filtering, reporting.
Attestations, SLA, dedicated instance, procurement support.
Carina gets people working. Scout makes that work safe to ship, share, and sell. The hook is the workspace. The moat is the security layer.
SCOUT_KEY. One line. Done.When a governed agent drifts beyond policy, behaves unexpectedly, or encounters a security event, Scout gives you one action: stop everything, from one place, in one second.
Every agent under Scout management receives a continuously updated trust profile covering identity, compliance, behaviour history, and risk score. Evidence. Not promises.
Verified identity record linked to deployment metadata, ownership, and capabilities.
Trust score updates in real time based on observed actions, policy compliance, and anomaly history.
Every action, approval, and governance event recorded and exportable as JSON, CSV, or PDF.
Three independent protection layers evaluate every input, tool response, and model context before any action is executed. Malicious prompts are blocked before they become behaviour.
Regex pre-filter and pattern matching on all user input and injected content before it reaches the model.
LLM judge layer evaluates context against active governance policy before approving execution.
Output guard inspects model responses and tool calls for unsafe, unexpected, or policy-violating content.
Monitoring tells you what happened. Scout decides what happens next.
| Traditional Security | Scout |
|---|---|
| Alerts | Intervention |
| Monitoring | Governance |
| Logging | Decisions |
| Reports | Control |
| Detection | Prevention |
Before agents take high-risk actions, Scout pauses and waits. Operators approve or reject from the dashboard.
Tool call or external action triggers policy evaluation.
Policy engine compares action against active governance template and risk thresholds.
High-risk actions pause. Human approves from the dashboard. Agent proceeds only on approval.
One Scout licence per deployment. Scout Monitor: observe. Scout Govern: act. Enterprise: comply.
Console URL and API key emailed at provisioning.
On Carina Cloud (£20/mo), add Scout Monitor at £49/mo or Scout Govern at £79/mo. Carina pricing
Add audit, trust, and response controls when the same workflow needs real guardrails.