Concepts
The three pages that make every other page read faster: how Sentisec works, the attack class it defends against, and what a verdict means.
Sentisec is a runtime control plane for AI agents. It sits between your agent and the model provider, evaluates every call in the context of the session it belongs to, and returns a verdict you can act on. Read these three pages before you wire anything — they take about ten minutes and everything downstream assumes them.
How Sentisec works
The request loop end to end: what you change, what the engine evaluates, what a block looks like on the wire, and where the record lands.
Threat model
The one failure class Sentisec is built for — an agent acting on instructions that arrived through its own context — and what it does not cover.
Verdicts and interventions
The four verdicts, the four interventions, and how to detect a stopped call from your code.
Verdict quick reference
NOMINAL nothing off; the steady state. ELEVATED anomalous; recorded for review. SUSPICIOUS off-policy; usually stopped. HIJACK high-confidence compromise; stopped and flagged.
Related
- Quickstart — wire it up once these make sense
- Comparison — how this differs from a gateway, a filter, and a WAF
- Security — how Sentisec handles your credentials and your data
n8n Gmail triage
Guard an inbox-triage automation so a downstream send never fires on attacker-authored instructions.
How Sentisec works
Point your agent at a Sentisec base URL. Supported model calls carry your workspace policy; tool proposals are checked before release, and evidence explains the decision.