Sentinel reads intent, not just keywords — surfacing targeted phishing, business email compromise and spoofing, then correlating them into campaigns and acting before they reach a person.
Rules and reputation lists stop the obvious. Sentinel is designed for the targeted, low-volume attacks that are written to look legitimate.
Multi-signal analysis weighs headers, payloads, sender history and intent — not just blocklists.
A layered pre-filter cascade clears most mail cheaply, sending only a small fraction to a model — fast model first.
Related messages are fingerprinted and grouped into scored campaigns — not treated as isolated emails.
Data minimisation, lawful-basis retention and one-command GDPR export or erasure.
Every message runs through specialised agents that each add a signal, then a correlation layer connects the dots across your mail flow.
Ingestion, headers, payloads, senders and intent — each handled by a dedicated agent.
Cheap, fast layers clear the obvious so models only see what genuinely needs them.
Most analysis runs on a fast, low-cost model; heavier reasoning is reserved for the few cases that warrant it.
Content fingerprinting plus infrastructure and timing signals group attacks into scored campaigns.
A clear customer view with on-demand Intelligence Reports generated without sending data to a model.
Auto-rules never punish an impersonated domain — generation is gated on authentication alignment.
Builds reputation over time so a familiar sender's sudden change of behaviour stands out.
DKIM, SPF and sender-IP extraction expose forged routing and authentication failures.
Isolated tenants and a clean API make Sentinel straightforward to run for one team or many.
Capture mail from your mailbox over a secure connection.
The cascade clears the obvious, escalating only what looks suspicious.
Agents inspect headers, payloads, senders and the message's intent.
Related messages are grouped into a campaign with a confidence score.
Protective rules apply automatically and a plain Intelligence Report is surfaced.
Sentinel's customer reports are generated from your own analysed data — the report layer runs without sending content to a language model, so insight doesn't come at the cost of privacy.
// data minimised · lawful-basis retention · audit-loggedStart small and move up as your mail volume and team grow. Every plan runs the same detection engine; higher tiers unlock more mailboxes, deeper analysis and richer reporting.
Put Sentinel in front of your mail and find out which "ordinary" emails were anything but.
Get started →