All systems observing Mountain View, CA— —:— —:— — UTC
Home / Products / Sentinel P.02 — Product

Sentinel

Catch sensitive data before it leaves.

Sentinel inspects prompts and payloads in-line, at capture time. It detects PII, credentials, source code, and confidential material — and redacts, hashes, or blocks it before it reaches an external model.

$ sentinel policy show --class credentials → action: block
Specification

Sentinel, at a glance.

Detector classesPII · credentials · source code · financial
Custom detectorsorganization-defined patterns & classifiers
Action modesredact · hash · block · alert
Enforcement pointin-line, at capture
Coveragechat · API · agent traffic
Incident routingalerting to security tooling
Capabilities

What Sentinel does.

In-line detection

Inspection happens in the request path, not after the fact. Sensitive content is handled before it leaves your boundary.

Redaction at capture

Configured classes are redacted or hashed at the moment of logging — the raw material never persists.

Custom detectors

Define patterns for internal identifiers, project codenames, or anything specific to your organization.

Incident alerts

Blocked or flagged events route to your security team with full event context from Ledger.

System

Works as one layer.

Sentinel shares one event record, one policy engine, and one dashboard with the rest of the platform. Each product is stronger for the others.

Early access

See Sentinel on your traffic.

Early-access deployments are supported directly by our team, with configuration matched to your environment.