Skip to main content
q-ai is an offensive security tool. It scans servers for vulnerabilities, generates adversarial payloads, intercepts traffic, tests AI agent susceptibility to injection attacks, and tracks execution via out-of-band callbacks. These capabilities are powerful and potentially disruptive.

​
Authorized Testing Only

Only test systems you own or have explicit written authorization to test. This means:
  • MCP servers you built and control
  • AI agent pipelines and RAG systems you built and control
  • Servers and targets in lab environments you operate
  • Third-party systems where you hold a written testing agreement or bug bounty scope that explicitly covers them
Running q-ai against a system without authorization may violate the Computer Fraud and Abuse Act (US), the Computer Misuse Act (UK), and equivalent laws in other jurisdictions. Unauthorized testing is illegal regardless of intent.

​
What Authorized Looks Like

If you are unsure whether you have authorization, you do not have authorization. Authorization requires explicit, documented permission — not implied permission, not assumed permission because you have credentials or API access, and not retroactive permission after testing has begun. For bug bounty programs, verify that MCP servers, AI agent infrastructure, and document ingestion pipelines are explicitly in scope before running any active scans, injection tests, or payload deployments.

​
Dangerous Payloads

The --dangerous flag enables IPI payload types that go beyond callback verification — data exfiltration, SSRF, and behavior modification. These payloads cause target systems to take real actions with real consequences. Use dangerous payloads only in isolated test environments where you fully control the blast radius. Verify that no sensitive data is at risk before deploying exfiltration payloads.

​
Responsible Disclosure

If you use q-ai to find a genuine vulnerability in a third-party system and have authorization to test, follow responsible disclosure:
  1. Notify the vendor before publishing — provide reproduction steps, evidence, and affected versions
  2. Allow reasonable time to patch — 90 days is the standard baseline; critical infrastructure may warrant an extension
  3. Coordinate publication — publish technical details after the vendor has had opportunity to respond, whether or not they do
For vulnerabilities in q-ai itself, see SECURITY.md.

​
Intended Use

q-ai is a security testing tool analogous to Burp Suite, Metasploit, or Nmap — purpose-built for authorized penetration testing and security research. It is not a general-purpose automation tool and should not be used to attack systems outside a controlled testing context. Use findings to improve the security posture of the systems you are responsible for protecting.