Build, govern and prove every agent.
ZAK is built on three primitives — so security teams get autonomy without losing control or auditability.
Drag triggers. Compose actions. Ship in minutes.
Every step an agent takes is written to the ledger — reason, decide, act, with a human in the loop wherever you want one.
Triggers
Actions
A reasoning loop, wrapped in governance.
Every ZAK agent runs the same loop — observe, reason, decide, act, verify — but each step passes through a policy gate, optional human approval, and the ledger.
Govern agents as code.
A declarative DSL defines exactly what an agent may do, when it needs approval, and what it can never touch — version-controlled and dry-run before it ships.
- Conditions on severity, exposure and CVaR impact
- Human-in-the-loop gates per action
- Hard guardrails & blast-radius limits
- Dry-run, version control, instant rollback
agent "vuln-remediation" { on: finding.created when: severity >= high and asset.exposure == "internet" and cvar_delta > 500_000 approve: slack("#sec-ops") # HITL gate do: patch(asset, via="crowdstrike") ticket("jira", project="SECOPS") never: touch(tag="production-db") ledger: true }
Six agents that ship with ZAK.
Pre-built, governed and ready to deploy — extend them or write your own in the Builder.
Autonomy you can prove.
Every agent action — and the reasoning behind it — is written to a tamper-evident ledger and mapped to the Secure Controls Framework. When the auditor asks "who did what, and why?", the answer is one query away.
- Tamper-evident, append-only audit trail
- SCF-mapped to your control frameworks
- Human-in-the-loop approvals where it matters
- Dry-run and version control before anything ships
Agentic security, answered.
What AI security agents are, how ZAK keeps them governed, and how you prove what they did.
Put agents to work — under your control.
Join the early-access program and bring governed, auditable automation to your security operations.
