ARGUS INTELLIGENCE

SENTINEL — the behavior scanner for AI-agent fleets

Send us your agent log lines. Get back a verdict on every event: a risk score, a component breakdown, which of 14 measured behavior classes matched, and hit counts. Your fleet, scored against the world's most complete library of real rogue-agent incidents.

The instrument was built from public post-incident disclosures across multiple AI systems — read end to end, correlated, and turned into a detection engine. Every finding we release is a conclusion from that evidence.

How it works

  1. Create a tenant — one API call, get a key. The free tier scores your first 1,000 events, no card.
  2. Stream log lines — POST /platform/ingest with your X-API-Key. Text goes in, verdicts come out.
  3. Read the verdict — score, component breakdown, behavior-class names, hit counts. Flagged events surface in your alerts feed.

Your telemetry stays in your tenant — it never enters our research store and is never searched by anyone but you. Your stored events are yours to erase at any time, one call, gone. And the ruleset never leaves our servers: customers receive verdicts, not our method.

Pricing

WATCH

$49/mo

Up to 1M scored events a month. Daily verdict rollups and behavior-class alerts. The wedge: try it without a purchase order.

Subscribe

SENTINEL

$249/mo

10M events a month. Per-agent scoreboards, incident digests, API + webhook alerts, 90-day verdict history. The workhorse.

Subscribe

FOUNDRY

$1,200/mo

Volume plus white-glove: custom behavior-class briefings, a quarterly fleet risk assessment, SSO, shared channel support.

Subscribe

Start now

curl -X POST https://sentinelscanner.io/platform/signup -H 'Content-Type: application/json' -d '{"email":"[email protected]"}'

Then score your first event: curl -X POST https://sentinelscanner.io/platform/ingest -H 'X-API-Key: YOUR_KEY' -H 'Content-Type: application/json' -d '{"events":[{"event_id":"e1","agent_name":"worker-7","text":"... log line ..."}]}'