The question nobody asks in the demo
Every testing tool now says "AI-powered." Very few answer the follow-up: when your AI analyzes my tests, where does my data go? For many tools the honest answer is "to a third-party LLM API", meaning your selectors, your test data, and fragments of your application's behavior leave your perimeter on every analysis.
If your security team would object to pasting source code into a public chatbot, it should ask the same question of testing tools. Most don't volunteer the answer.
Deterministic and local, on purpose
Our intelligence (smart test selection, self-healing proposals, defect and coverage analysis) is deterministic and runs inside your tenant. No code, selector, or test data is sent to a third-party model by default. The single optional LLM path in the platform (LLM-assisted API testing) is bring-your-own-key: your key, your provider agreement, your explicit opt-in.
This also makes behavior reproducible: the same input produces the same analysis, which matters when a test-selection decision needs to be audited six months later. Try auditing a probabilistic API call you don't control.
Three questions for any AI-testing vendor
- Exactly which data leaves our environment, to whom, and under what agreement?
- Can the AI features run with external calls disabled, and what still works?
- Are AI-driven decisions reproducible and auditable after the fact?
How we measured this
- Model calls from test intelligence go to the local model runtime inside the deployment. The single outbound path is the bring-your-own-key resolver, which stays inactive until a tenant configures a key. Both are inspectable in the codebase and in the network egress of a running tenant.