Upload a CSV. Get ranked, risk-scored cost fixes as sprint tickets.
Isolation Forest scores every line item. Rows above threshold match against a per-provider pattern library.
Risk and failure-cost reconciled through the tier safety model. Protected services flagged, never silently dropped.
Ranked by ROI, each ticket carries a savings range and effort estimate. Drops straight into a sprint.
3 of 8 instances idle above 40% for 14 days. Right-size m5.2xlarge → m5.xlarge and enable target-tracking on the ASG. No tier-1 or tier-2 dependencies.
2.4 TB in Standard untouched for 90+ days. Add a lifecycle rule to Glacier Instant Retrieval. Read pattern is archival, not hot.
Node type looks over-provisioned, but this is a tier-2 core service. Surfaced for manual review and sorted to the bottom of the ranked list — never applied automatically.
No. CSV upload only — there's no live billing API integration, no IAM role to grant, no read access to your account. You export a billing CSV when you want an audit and upload it. Nothing runs against your infrastructure.
Dropped from recommendations entirely. You mark them tier-1 during setup; the conflict resolver removes them before ranking, so they never appear in a ticket. Protected means untouched.
It's flagged as conflicted and surfaced, not hidden. Any tier-2 touch or risk step-up gets marked with its reason and forced to the bottom of the ranked list. You decide — the tool never quietly buries a risky change.
Row-level security on every table, scoped to your user id. Backend writes still independently verify ownership per request, and a mismatch returns 404 — a request can't even tell whether someone else's row exists. Single-tenant throughout.
Yes — 30-day retention. A nightly job hard-deletes anything past its expiry, and soft-delete is available anytime you want a report gone sooner. The PDF export stays downloadable across that window.