90-second product walkthrough
See the path from cluster overview to exact application context, evidence-backed investigation, review-first fix and optimization.
Start the walkthrough →
KrevoPilot
Talk to the team
Four concise, versioned resources explain what the product does, what the agent can access, what data leaves a cluster and what our published evidence proves.
See the path from cluster overview to exact application context, evidence-backed investigation, review-first fix and optimization.
Start the walkthrough →Read the dated scenario corpus, scoring method, exact versions, result and explicit limitations behind the accuracy claim.
Review benchmark evidence →A printable answer to installation, connectivity, data handling, authentication, tenant access, audit and retention questions.
Download one-page PDF →Inspect installed Kubernetes objects, exact resources and verbs, conditional collectors, Secret handling and runtime hardening.
Review agent permissions →An in-cluster Deployment, dedicated ServiceAccount, read-only ClusterRole/Binding, connection Secret and optional egress NetworkPolicy.
No. The published collection role contains no create, update, patch or delete verbs. Remediation is review-first.
Bounded health/state, metrics, selected events, configured manifest summaries and optional limited log evidence.
Secret values are never transmitted. When manifest collection is enabled, existence, type and key names are summarized so broken references can be diagnosed without exposing values.
Password sign-in and workspace OIDC/SSO are supported. Roles, teams and cluster assignment control access.
Security-sensitive and administrative actions retain actor attribution for workspace audit review.
The benchmark is a bounded controlled-corpus result—not a universal accuracy guarantee. This package describes implemented technical controls, not a certification or contractual DPA.
Read the full trust report →