From a Simple Objective
to a Complete Execution Record
You tell Zarb what to test. It builds the plan, runs the commands safely in your dedicated environment, and logs exactly what happened at every single step.
01 / PLANNING
Zarb turns your request into a clear plan.
Zarb takes your plain-English objective, gathers context from any URLs or attached files, and determines the best methodology. The result is an ordered list of commands, with a documented reason attached to every single one.
- Natural-language objectives
- Methodology skills
- URLs and attached files as context
- Commands and rationale stored together
02 / REVIEW
You decide how much control you want.
Missions pause after planning so you can review the proposed commands before starting, or even execute them one step at a time. Scans are designed to be hands-off and move into execution automatically. In either mode, you can pause, resume, or cancel individual steps at any time.
- Missions wait for Start
- Scans advance automatically
- Single-step execution for idle Missions
- Pause, Resume, and Cancel controls
03 / EXECUTION
Tests run safely in your dedicated Alpha environment.
An Alpha is an execution environment assigned exclusively to your organization. Zarb can manage it for you, or you can connect your own infrastructure over direct SSH. Output streams back live while a step is running, and up to four steps can execute concurrently to save time.
- Managed or customer-owned Alpha
- Connected over direct SSH
- Live output streaming
- Bounded per-step jobs
04 / ADAPTATION
The AI adapts based on real-time results.
After every step, the agent analyzes the output it just produced. It can generate follow-up commands, remove steps that no longer make sense, or pick the best path forward. Findings are logged and refined immediately, meaning evidence is built during the run, not just assembled at the end.
- Post-step analysis
- Follow-up creation and retirement
- Platform-owned ordering
- Findings recorded during execution
05 / OUTPUTS
You get a comprehensive, evidence-backed report.
Zarb automatically generates a high-level executive summary and a granular technical report for every run. Both are downloadable as PDFs and are written directly from the stored evidence.
- Executive and detailed reports
- Both downloadable as PDFs
- Written from the stored evidence
- Phase-level and mission-level diagrams
Four Kinds of Limit, Kept Apart
Zarb enforces four distinct kinds of limit. Some are architectural, some apply only to scans, some are controls you hold, and some are rules we operate by. They are worth separating, because they are not equally strong.
Architecture
Commands execute on an Alpha belonging to a single organization, and records in Zarb are isolated by organization.
Scan controls
Web scans use declared origins and network scans use declared hosts. Anything outside that is rejected from the coverage ledger.
Operator controls
Missions wait for you to press Start. Active runs can be paused and running steps can be cancelled.
Operating rules
You must be authorized to test the target. The methodologies are strictly non-destructive and exclude Denial of Service (DoS) attacks and pivoting to third parties.