Nothing leaves. That's the architecture, not a setting.
Krenox deploys entirely inside your AWS account. Telemetry, code context, incident memory, and AI inference all stay behind your boundary. There is no toggle to get this wrong.
Read-only by construction
The IAM role Krenox runs under contains no write actions; you can audit the policy before you grant it. Collectors observe; they never mutate. The only artifacts Krenox produces are findings in your account and pull requests in your repository.
Inference runs against Bedrock in your account: you choose the model, your data-use agreement with AWS applies, and prompts containing your telemetry never transit our systems. Inference cost lands on your AWS bill, visibly, not hidden in ours.
Krenox holds no credentials to your runtime. Its one write path is a pull request your team reviews and your CI/CD ships. Reviewable, auditable, reversible: the same guarantees you demand of every engineer.
Your incidents, findings, and reports download from your dashboards. Uninstall is one command. There is no data hostage scenario, because your data never left your account.
No cross-customer learning. Ever.
What Krenox learns from your incidents stays in your account and improves only your diagnoses. Nothing about your systems (not patterns, not embeddings, not "anonymized aggregates") feeds anyone else's deployment. Your operational history is a competitive asset; we treat it that way.
Ask us the hard questions
We're an early-stage product and we'll answer your security questionnaire directly, including the places where the honest answer is "not yet." The architecture above is the part we will never compromise on; certifications and audits follow from it, on request we'll share where each one stands.
Send us your questionnaire →