Offensive AI security,
built by operators.
KageX is an offensive AI security platform for people who want to test, break and secure real-world AI systems: hands-on labs, a structured curriculum, free challenges and open-source tools, all built around how AI actually fails in production.
- FreakLabs — paid hands-on labs and a 9-level course
- SentinelGoat — free progressive AI hacking challenges
- AgentBreaker — open-source automated AI testing
- Mirage — open-source browser extension that red-teams your AI
Everything you need to learn AI security
Practice on real AI systems, challenge yourself with free CTFs, and automate testing with open-source tools.
OWASP LLM Top 10 Coverage
FreakLabs and AgentBreaker map directly to 6 of 10 OWASP LLM categories. Here is exactly what is covered — and what is coming next.
LLM01
Prompt Injection
CoveredLLM02
Insecure Output
CoveredLLM03
Training Data Poisoning
CoveredLLM04
Model Denial of Service
RoadmapLLM05
Supply Chain Vulns
RoadmapLLM06
Sensitive Info Disclosure
CoveredLLM07
Insecure Plugin Design
CoveredLLM08
Excessive Agency
CoveredLLM09
Overreliance
RoadmapLLM10
Model Theft
RoadmapHow it works
Learn the theory
Start with structured lessons that explain how AI gets attacked — from basic concepts to advanced techniques.
Academy
Practice on real systems
Launch a private lab and try the attack yourself. Built-in hints if you get stuck. Mark it complete when you're done.
Labs
Earn your certificate
Complete all labs and get a verifiable AI Red Team Operator certificate you can share on LinkedIn.
Certificate
Frequently asked questions
What is KageX?
KageX is an offense-first AI security company. It builds hands-on training and open-source tooling for testing how AI systems fail, covering prompt injection, retrieval and agent attacks, and LLM red teaming. The platform has four products: FreakLabs, SentinelGoat, AgentBreaker and Mirage.
What is the difference between FreakLabs, SentinelGoat, AgentBreaker and Mirage?
FreakLabs is the hosted, paid training product: 20 attack labs against real models and a 9-level curriculum that ends in a verifiable certificate. SentinelGoat is a free AI security CTF with leaderboards. AgentBreaker is a free, open-source tool that runs adversarial tests against an LLM app or agent and produces a vulnerability report. Mirage is a free, open-source browser extension that runs an autonomous red-team agent against any AI app with deterministic scoring.
Is KageX free to use?
Mostly. SentinelGoat, AgentBreaker and Mirage are free, and the last two are open source. FreakLabs is paid, with a free tier: two labs and the first two course levels are free and need no credit card.
Who is KageX for?
Red teams, penetration testers, developers shipping AI features, and security researchers. It suits anyone who needs to learn how AI systems break by attacking them, and teams that need a repeatable way to test their own LLM apps and agents.
Which security frameworks does KageX map to?
The OWASP Top 10 for LLM Applications and MITRE ATLAS. Labs, challenges and tool reports use those as a shared taxonomy, so findings line up with the frameworks security teams already report against.
What is the AI Red Team Operator certificate?
The credential earned by completing all 20 FreakLabs labs. It is verifiable, with a unique ID anyone can check, and can be shared on LinkedIn.
Ready to start?
2 labs and 2 course levels are completely free. No credit card required.