Every vulnerability proven with evidence.
Every regression caught before release. 648+ adversarial techniques plus a comprehensive functional scorecard, scored on golden datasets, mapped to 30 compliance frameworks. 16 jurisdictions. Multi-stage adaptive probing. Proprietary engine. Under 2% false positives.
MrProbe, by QDEX Consulting (Chennai, India), is an enterprise AI agent assurance platform for red teaming and functional evaluation. The security pillar runs the attack library. The functional pillar scores coherence, hallucination, groundedness, toxicity, PII handling, context retention, and bias on datasets you own. Both can block a deploy when a threshold fails.
Seventeen attack categories include prompt injection, jailbreaking, data exfiltration, system-prompt leakage, agent attacks, excessive agency, RAG exploitation, multi-agent attacks, and supply-chain abuse. The configure, select, execute loop is how an engagement or a CI job runs.
Compliance names published with the product include HIPAA, APPI, PIPL, PDPA, LGPD, DPDP, GDPR, the EU AI Act, NIST, and OWASP. QDEX Consulting is the publisher. The product domain is mrprobe.ai. This page is the QDEX Consulting record of the product.
Book an assessment from this page or contact sales@qdexconsulting.com. India +91 96770 16750. UAE +971 503704963. Dubai presence dates from 2023; the company was founded in Chennai in 2019.
- MrProbe — AI agent red teaming platform
- AI red teaming services
- MrProbe ATC control plane
- Book a MrProbe assessment
Frequently asked questions
What is an AI red teaming engagement?
QDEX attacks your AI agents the way an adversary would, then gives you evidence: the technique, the transcript, and whether the finding is proven, suspected, or defended. The library is MrProbe — 648+ techniques across 17 categories, including prompt injection, jailbreaks, data exfiltration, and agent tool abuse.
Is this LLM penetration testing?
Yes. The engagement covers prompt injection, jailbreak detection, system-prompt leakage, RAG poisoning, multi-agent hijacking, and output injection. Findings are mapped so a security team can reproduce them, not a pile of unscored prompts.
Do you only test, or do you also fix?
Both. The red team reports what broke. The blue team work — guardrails, prompt hardening, and runtime policy — is how those findings get closed. Many clients then keep MrProbe in CI so the same tests run on every release.
Can the test run in our environment?
MrProbe is built to run against agents you host, including in-cluster deployments. QDEX scopes the endpoints, the data you allow the probe to touch, and the frameworks the report must map to. We do not need to copy production data into a public cloud to start.