# ZopNight v2.0: The Control Layer Your Cloud Bill Has Been Missing > Most teams treat cloud waste as a reporting problem. It isn't. ZopNight v2.0 ships the full four-layer control stack: discovery, policy, audit, and action — across AWS, GCP, and Azure. Source: https://zop.dev/resources/blogs/cloud-bill-is-a-control-problem Published: 2026-04-24 · Author: riya-mittal · Tags: cloud, finops, zopnight, cloud-governance, cloudops --- We've been watching cloud bills grow for years. Dashboards got prettier. Alerts got louder. The bills kept climbing. ZopNight v2.0 is our answer to why: the problem was never visibility. It was control. This release ships the full four-layer stack every multi-cloud team needs: discovery with 14-day metrics, policy that binds, audit that remembers, and action that executes. One platform, three clouds, no agent, 60-second connect. ## The problem: observation is not control Industry surveys have put average cloud waste at 32% of spend for years running. That number does not move because the default tooling is reporting tools, not control tools. A dashboard ends in a human. The human files a ticket. The ticket joins a queue. A control loop closes automatically: evidence triggers policy, policy triggers action, action writes to audit. That is the entire difference. ![Observation vs Control Loop](https://storage.googleapis.com/zopdev-blog-resources/1/files/originals/20260427/8a491b12-e376-4a46-b49d-ed88cd98b049-diagram1.png) ZopNight v2.0 closes that loop across AWS, GCP, and Azure without an agent in your cluster. ## Feature 1: Schedule anything — no cron required The most common non-prod waste driver is resources running when nobody is using them. ZopNight's scheduler is a 7-day visual grid. Green means on, red means off. You drag to set the window. ZopNight generates the cron, evaluates per minute, and fires idempotent actions. [Dependency awareness](https://zop.dev/resources/blogs/schedule-override-the-safety-valve-your-cloud-automation-has-been-missing) is what makes this production-safe. Real environments have boot order requirements. Resource groups encode that order — ZopNight starts in sequence, tears down in reverse. No script can replicate this safely. [Cron can't do it at all](https://zop.dev/resources/blogs/cron-vs-purpose-built-scheduler-the-hidden-maintenance-tax-on-cloud-cost-optimization). ![Schedule Engine and Resource Group Boot Order](https://storage.googleapis.com/zopdev-blog-resources/1/files/originals/20260427/aa008722-94f3-4c50-92e9-fd00e52fa9ec-diagram2.png) Every override has a start, an `expiresAt`, a reason, and an auditable owner. Maintenance windows that break schedules by design get recorded, not silently bypassed. ## Feature 2: 337 audit rules — every one shows its evidence "You could save 25%" gets ignored. "Your `m5.xlarge` in `us-east-1` ran at 8% p95 CPU for 14 days; moving to `m5.large` saves $93 per month with 96% confidence" gets acted on. ZopNight v2.0 ships 337 pre-built rules: AWS 155, GCP 75, Azure 107. Seven categories: idle, rightsizing, schedule, orphan, compliance, discount, governance. Every rule carries its evidence window, its metric, its current cost, and its optimised cost. The recommendation shows its work. ![337 Audit Rules — Evidence to Action Pipeline](https://storage.googleapis.com/zopdev-blog-resources/1/files/originals/20260427/7855c85e-786f-42a5-9af8-0555662a19c5-diagram3.png) Discovery is topology-aware: EKS clusters surface with their node groups, Databricks with instance pools, managed databases with replicas. Right-sizing calls have the full tree. That is why the recommendations are [closed-loop, not speculative](https://zop.dev/resources/blogs/closed-loop-cloud-remediation). ## Feature 3: Auto-tag at discovery — not at cleanup [Tag governance enforced at discovery time](https://zop.dev/resources/blogs/stop-doing-tag-cleanups-start-doing-tag-governance-at-discovery-time) holds untagged resource rates below 5%. Quarterly cleanup campaigns typically stay above 20% and rising. The difference is timing: cleanup fights existing drift, governance prevents it. ZopNight predicts `prod`, `staging`, `dev`, and `noStop` the moment a resource appears. A 1-100 confidence score is assigned and held for human approval before anything persists. The approved tag feeds both scheduling and showback — engineering and finance use the same taxonomy automatically. | Approach | Untagged rate | Finance reconciliation | Cleanup cadence | |---|---|---|---| | Quarterly cleanup | 20%+ | Manual spreadsheet | Every 3 months | | Discovery-time governance | Below 5% | Automatic showback | Never needed | ## Feature 4: MCP server — AI that queries live state Engineers waste 45 minutes reconstructing what happened to a cloud resource. They screenshot a dashboard, paste it into ChatGPT, describe the error, wait. The answer is as stale as whatever they pasted. ZopNight v2.0 ships an MCP server with 43 read-only tools over streamable HTTP. Cursor, Claude Code, Codex, and Windsurf connect with a PAT and query live governance context directly. "Which schedules fired in the last two hours and what did they change?" returns a live answer, not a reconstruction. ![MCP Server — AI Tools to Live Governance State](https://storage.googleapis.com/zopdev-blog-resources/1/files/originals/20260427/0b45c7ad-fb02-4041-8176-94fb036193a6-diagram4.png) Also available via 125+ REST endpoints with consistent pagination and retry semantics. PATs support flexible expiry. OIDC covers Azure federation. ## Feature 5: Atlas — your infrastructure on a globe ZopNight's Atlas view renders your infrastructure as a 3D globe with inter-region arcs for VPC peering, transit gateways, and interconnects. Drill from provider to node group on a single canvas. Cross-region blast radius is visible in one click instead of three spreadsheets. Before you enforce a schedule or apply a right-sizing rule, Atlas shows you where that policy lands. Governance without topology is guesswork. ZopNight Atlas — 40 regions, 5,160 resources across AWS, Azure, and GCP on a single globe ## Feature 6: Cost reporting — one screen, actionable Six dashboard widgets. Cost forecasting. Showback by team or tag with reconciliation. Multi-currency. Budget health. Anomaly detection with root-cause breakdown. CSV exports that match what is on screen. Finance stops rebuilding your report. The labels engineering sets at discovery time are the same labels finance reports against. ## Feature 7: RBAC and audit — governance without gatekeeping The [cloud governance RBAC model](https://zop.dev/resources/blogs/cloud-governance-rbac-viewer-editor-admin-custom-roles) in ZopNight v2.0 is built around an operational reality: the person who needs to see history is usually not the person who needs to change policy. Three tiers: Viewer with 16 policy categories, Editor with 32, Admin with 52. Custom roles fill the gaps. Every start, stop, tag change, and override writes a full audit record: resource, status, source, triggered by, timestamp with timezone. Filterable across all five dimensions. ![RBAC Roles, Audit Records, and Notifications](https://storage.googleapis.com/zopdev-blog-resources/1/files/originals/20260427/300d01a7-30ed-4164-af2f-ad1a5f1100b0-diagram5.png) When an auditor asks who turned that on, when, and why, the answer is one filter away — not a Slack thread and three screenshots. ## Try it without talking to anyone The [Playground](https://zop.dev/app/redirect?playground=true) is loaded with finance, retail, and healthcare datasets. Same UI as production. Read-only, no persistence, no auth required. You get the full ZopNight experience — schedules, audit rules, topology, tagging — against realistic data before you commit to anything. If you want the architecture behind all seven features, the [v2.0 deep dive](https://zop.dev/resources/blogs/zopnight-v2-deep-dive) covers the design decisions behind each layer. The goal was never a prettier chart. It was a bill that matches what your team actually intended to run. ZopNight v2.0 is the mechanism. Everything else is noise.