Product / How It Works

What happens between the alert and the answer.

Most platforms show you the alert and the answer and leave the middle to you. Trace8 is the middle — a multi-agent pipeline that investigates, contests its own findings, and hands you a decision with its work shown.

The pipeline, in full

Each agent: what it does, what it reads, what it outputs.

Triagethe opening read
Ingests the raw alert, normalizes it, returns a structured verdict: true positive, false positive, or escalate — with severity and reasoning. Structured JSON via tool-use, so every downstream agent consumes a clean contract, not free text.
Huntcorroborate and contradict
Pivots across available log sources for evidence on both sides. Hunt is not trying to confirm Triage — it’s gathering everything relevant, including the facts that argue against the initial call.
Challengethe adversary
The architectural commitment. Challenge takes the Triage verdict and attempts to dismantle it: what would have to be true for this to be a false positive? What benign explanation fits the same evidence? A verdict that survives Challenge is one you can act on. A verdict that doesn’t gets flagged, not buried.
Evidencethe case file
Separates the known from the assumed from the missing. This is what makes the output auditable: an analyst can see not just the conclusion but the gaps in it.
EDR Managerhost-level ground truth
Reaches into endpoint telemetry to confirm or deny what the logs suggest. Process lineage, parent-child relationships, signing status.
Email Guardianthe delivery vector
For anything email-borne: resolves sender reputation, payload, and the full delivery path. Closes the “how did this get in” question.
Brieferthe human handoff
Writes the narrative a human actually reads. Not a field dump — a decision, the reasoning behind it, and what to do next.
Auditorthe final check
Validates the entire chain before it surfaces. Catches internal contradictions, broken reasoning, incomplete evidence. The last gate.

Why structured output matters

A reliability guarantee, not an implementation detail.

Every agent emits structured JSON through tool-use, not prose. Free-text handoffs between agents accumulate ambiguity at every step; structured contracts don’t. It’s the difference between a pipeline you can audit and a chain of suggestions you have to babysit.

The data flywheel

Corrections become capability.

When an analyst overrides a verdict, that correction is captured and fed back into the improvement loop. Not logged and forgotten — used. The agents that investigate your alerts in month six have learned from every correction made in months one through five.

Accuracy isn’t a fixed spec. It’s a curve that bends in your favor.

Connectors

Add a source by dropping a file.

New connector, no code change. Connector definitions are JSON — drop the file, the source is live. The catalog grows without an engineering cycle, which means your coverage isn’t gated on our roadmap.

See it run against your data.

A working pipeline, your alerts, the full eight-agent chain.