Agent monitoring and risk control

Watch agents. Stop bad runs. Prove what happened.

Scout monitors agent sessions, raises alerts, and gives operators a fast way to respond. It does not run tasks; your agent does. Carina is the reference agent; any stack can integrate via HTTP or SDK.

Public trust API Carina (does the work) Integrate any agent

Monitor live sessions Respond to risky runs Report outcomes clearly
scout-console; example: carina-prod-01
14:02:11DECEPTIONMarker rotated · decoy credentials refreshed
14:02:44INJECTIONSuspicious instruction in tool response · blocked
14:03:02ATTESTERC-8126 WAV queued · wallet required
14:03:15ALERTAnomaly score 92 · channel route matched
14:03:16KILL SWITCHAgent suspended · awaiting review
Threat level
Carina paired Yes
Events today 1,284
Production controls

What Scout does when an agent misbehaves

Not a chat UI. A control plane: see events, quarantine tools, alert your team, and halt execution.

Live monitoring

See active agent sessions, event volume, and risk movement in one place.

Incident detection

Scout highlights suspicious runs so operators can review them quickly.

Real-time security dashboard

Live event stream, activity summaries, and anomaly charts with operator review flows.

Response controls

Operators can pause or quarantine a run when they need to intervene.

Alerts where you work

Send critical notifications to the channels your team already uses.

Fast stop button

Pause a compromised run from the dashboard and review it before resuming.

Reporting

Keep an audit trail and export the records your team needs.

Policy guardrails

Apply response policies consistently across connected agents.

Verification

Detect incidents. Keep records.

Live risk metrics ship today at carinaai.uk/trust. When configured, Scout can publish public verification data for the systems you choose.

Type 1

Runtime issues

Scout identifies a run that needs attention and records enough context for review.

Type 2

Access anomalies

Scout surfaces unexpected access patterns so operators can respond quickly.

Type 3

Reasoning compromise

Scout highlights suspicious instructions or poisoned context and preserves the evidence.

Carina pairs in setup: run carina setup with credentials from your provisioning email. Other agents use the HTTP events API or SDKs. Public read-only metrics: carinaai.uk/trust.

Agent-agnostic

Scout watches the agent. Any agent.

Carina is the reference workforce agent and pairs in one setup wizard. Hermes, OpenClaw, AutoGPT, and custom stacks connect via HTTP events or SDK adapters.

OpenClaw
Hermes
AutoGPT
Custom agents

Carina product details (tools, packs, channels) live on carinaai.uk.

Pricing

Monitor or Pro

One Scout licence per deployment. Monitor gives visibility; Pro adds response controls and alerts. Carina Core is free on carinaai.uk.

Scout Monitor

£79/month
  • Live monitoring dashboard
  • Risk signals and anomaly detection
  • Audit history
  • Browser alerts (dashboard open)
Start Monitor →

Enterprise

Custom
  • Everything in Scout Pro
  • Verification options when configured
  • Custom reporting
  • 99.9% uptime SLA
  • Dedicated support
  • On-premise option
Contact sales →
Get started in minutes

Connect Scout in three steps

Deploy Scout beside your agent. Carina pairs in setup; other stacks use proxy mode, the HTTP events API, or SDKs.

1

Deploy Scout

Run the Scout service on your VPS alongside your agent.

SCOUT_API_KEY=... pnpm start
2

Connect agent

For Carina, run carina setup and paste your Scout API key. Other agents: use the HTTP events API or Node SDK from the docs.

LABYRINTH_ENABLED=true carina setup
3

Activate licence

Enter your licence key and any alert channel credentials. Scout begins monitoring immediately.

scout activate --key lscout_xxx

Need monitoring before you ship agents?

Scout Monitor is £79/mo; Scout Pro is £99/mo per deployment. Your private console URL and API key are emailed when provisioned. Carina installs separately and is free.

Get Scout Pro; £99/mo → Monitor £79/mo Carina agent (free)