Home GitHub
1 / 10

Autonomous Security Testing

Donna

AI-Powered Pentesting

Grab 'em by the balls.

13 AI agents. One mission. Find every vulnerability before attackers do.

2026

The Question

What if security testing
never sleeps?

Attackers don't wait

Automated bots scan the internet 24/7. New CVEs are weaponized within hours. Your security assessment from last quarter is already obsolete.

The talent gap is real

3.5 million unfilled cybersecurity positions globally. The few experts available are overwhelmed, overworked, and expensive.

AI changes the equation

Large language models can reason about code, craft payloads, and chain vulnerabilities — the same creative thinking that makes great pentesters, now available on demand.

The Problem with Pentesting

"Traditional penetration testing is slow, expensive, and inconsistent. A single engagement can take weeks, cost tens of thousands, and still miss critical vulnerabilities."
21 avg days per engagement
$25k+ typical cost per test
60% of vulns missed by manual testing
Broken shield illustration

Meet Donna

13 specialized AI agents, orchestrated to think like a hacker.

Donna

Recon

5 agents

Analysis

4 agents

Exploit

3 agents

Report

1 agent

The Pentesting Pipeline

Four phases. Fully automated. Durable execution.

1

Reconnaissance

Target enumeration, port scanning, technology detection, OSINT gathering

2

Analysis

Vulnerability scanning, attack surface mapping, weakness identification

3

Exploitation

Proof-of-concept validation, payload testing, privilege escalation

4

Reporting

Structured findings, severity ratings, remediation guidance, executive summary

How Agents Think

1Observe
2Plan
3Execute
4Verify
Repeat
"Each agent runs Claude in a sandboxed Docker container with controlled tools. No hallucinated findings — every vulnerability is verified."
ContainerNmap Scanner
ContainerWeb Crawler
ContainerSQL Injection
ContainerXSS Tester
Temporal workflow illustration

Durable Execution

Powered by Temporal — workflows survive crashes, network failures, and restarts. Every step is checkpointed.

Auto-retry on failure Resume from checkpoint Full audit trail Real-time observability

Real-Time Dashboard

Monitor every workflow. Track every finding. All in one place.

🛡Donna
3Running
47Completed
12Critical
TargetStatusFindingsDuration
api.example.comRunning3 Critical12m 47s
staging.app.ioCompleted5 High34m 12s
cdn.service.netCompleted2 Medium18m 03s

Structured Findings

Every vulnerability comes with evidence, severity rating, and actionable remediation steps.

CriticalSQL Injection in auth endpoint
HighStored XSS in user profiles
MediumMissing rate limiting on API
LowVerbose error messages exposed
InfoServer version header disclosed
CRITICAL CVSS 9.8

SQL Injection — /api/auth/login

Evidence 
POST /api/auth/login
{"email": "' OR 1=1--", "pass": "x"}

Response: 200 OK
{"token": "eyJhbG..."}
Remediation

Use parameterized queries. Validate and sanitize all input.

Secure your infrastructure.
Automatically.

Grab 'em by the balls.

13 AI Agents / Temporal Orchestration / Full Reports
View on GitHub
git clone https://github.com/schlunsen/donna.git