AI AGENT
SECURITY TESTING
METHOD / SOURCES / LIMITS

Know what a listing
is based on.

A provider’s public offering is evidence of what it says it sells. It is not evidence that we have tested its work.

How a provider enters the directory

We review an official service page for an offering relevant to AI agents, LLM applications or their integration boundaries. Firms that only use AI to test conventional software do not qualify on that basis alone. Hybrid firms are labelled so a platform and an assessment service are not confused.

Each published profile links to its source and identifies the fields supported by it. Records carry an explicit publication status. Research leads with unavailable or inadequate evidence remain drafts and do not generate profiles.

Confirm availability, capacity, commercial terms and your specific requirements directly with the provider before buying.

How we handle unknowns

“Not published” means a detail was not recorded in the reviewed source. “Confirm with provider” marks a procurement question. Neither label means the service is unavailable. We do not infer hosting options, certifications, pricing or retest terms from general marketing language.

Directory ordering

Providers are listed alphabetically.

Editorial recommendations and limits

“Relevant when” sections are editorial interpretations of public positioning. We publish no supplier scores, reviews or claims of independent testing. An assessment cannot guarantee security, and a directory listing cannot establish suitability for your system.

Corrections

Email contact@aiagentsecuritytesting.com with the profile URL, disputed statement and a relevant primary source. Please leave out confidential evidence.

Explore the researched providers →