AI investigation
AI investigation is a Pentestas scan mode for goal-driven testing. A planner coordinates workers, collects observations and builds an evidence graph inside the existing scan. It uses the AI connection configured in Settings.
Before starting
- Use an account on a plan that enables AI features.
- Configure Anthropic, DeepSeek or Grok in Settings → AI-Powered Scanning.
- Verify the target domain unless your account already has unrestricted scope.
Start an investigation
- Open New Scan and select a web application target.
- Enter the target URL.
- Under Advanced → Scan Mode, select AI investigation.
- Enter an Investigation goal, such as “Review authorization boundaries and collect reproducible evidence.”
- Start the scan. Open it from the existing Scans list.
For follow-up work, open an existing scan and select Investigate with AI. This creates a new scan with the original findings as planner context.
Use the Investigation tab
Open Scan detail → Investigation:
| Section | What you can do |
|---|---|
| Planner | Read the objective, send instructions, pause or resume the investigation |
| Goals | Review progress and add objectives |
| Constraints | Add operating boundaries for the planner and workers |
| Workers | Inspect assignments, pause, cancel or retry work, and send instructions |
| Approvals | Inspect a proposed tool action and allow or deny it |
| Evidence graph | Inspect how observations, goals and findings relate |
| Conversations and execution history | Read planner/worker messages and open full execution evidence |
Live Feed displays investigation activity. Conversation history supports older messages. There is no separate application or login.
Findings and reports
Investigation findings appear in the scan's Findings tab and the platform's global Findings page. Use the existing verification, triage, export and report workflows. An AI assertion is not independent proof: imported observations start unverified until the platform's validation workflow establishes otherwise.
Stop, Rescan and Delete use the existing scan controls. Deleting the scan also removes its underlying investigation task.
If AI investigation is unavailable
- Check that your plan enables AI and that an AI connection is saved in Settings.
- Resolve any target-domain verification or concurrent-scan-limit error.
- If the engine is unavailable, contact your platform administrator.
Saved browser sessions are not automatically transferred into investigation shell tasks. Operating constraints guide the agents; they do not constitute a network firewall. Administrators should consult the deployment documentation for execution boundaries and third-party license terms.