Nutanix AIops

> Disclaimer: Community-maintained open-source project, not affiliated with, endorsed by, or sponsored by Nutanix. Product and trademark names belong to their owners. Source at github.com/AIops-tools/Nutanix-AIops under the MIT license.

Governed Nutanix Prism Central (v4 REST API) operations — 51 MCP tools, every one wrapped with the bundled @governed_tool harness: a local unified audit log under ~/.nutanix-aiops/, token/runaway budget guard, undo-token recording, and descriptive risk-tier labels. The Prism Central password is stored encrypted (~/.nutanix-aiops/secrets.enc, Fernet + scrypt) — never plaintext on disk.

What sets it apart from read-only Nutanix MCPs: (1) automatic ETag / If-Match on every mutation — the v4 footgun handled for you; (2) automatic pagination; (3) mixed-hypervisor VM listing (AHV + ESXi, relevant to hypervisor-migration estates); and (4) the governance harness with dry-run + double-confirm on destructive writes.

> Standalone: the governance harness is bundled in the package (nutanix_aiops.governance) — no external skill-family dependency.

What This Skill Does

The CLI is a convenience subset; the full 51-tool surface is via the MCP server. See references/capabilities.md for the tool → API-path → returns map.

Quick Install

uv tool install nutanix-aiops
nutanix-aiops init       # interactive wizard: PC host/port 9440/username + encrypted password
nutanix-aiops doctor     # connectivity + REST-RBAC preflight

Or as an OpenClaw plugin, which installs this skill and its MCP server together:

openclaw plugins install clawhub:@zw008/nutanix-aiops
openclaw skills info nutanix-aiops          # expect: Visible to model: yes

Needs uvx on PATH: the MCP server is fetched with uv, pinned to this release.

When to Use This Skill

• Diagnose the estate in one shot (diagnose cluster-health): degraded resiliency, storage pools/containers over 80% / 90%, nodes down or missing — worst-first, each finding citing the measured number
• Triage the alert backlog (diagnose alert-triage): per-severity counts, unacknowledged criticals, the oldest unresolved alert and its age
• Inspect the estate (overview, cluster health, cluster util): clusters, hosts, resiliency, utilization
• VM lifecycle across AHV + ESXi (vm list/get/power/create/update/clone), and guarded destructive ops (vm delete, vm migrate) with dry-run + double-confirm
• Root-cause an alert (analyze_alert) — correlate it with related events into a probable-cause + suggested-actions summary
• Data protection: snapshots, recovery points, protection domains, vm_protect, pd_failover
• Upgrades (lcm_inventory → lcm_precheck → lcm_update) and capacity forecasting (capacity_runway)

Do NOT use when the target is a non-Nutanix platform — this skill is Prism Central v4 only. For other infrastructure, use the appropriate other AIops-tools sibling.

Common Workflows

"Something is wrong with the estate" → diagnose, then act

1. nutanix-aiops diagnose cluster-health → worst-first findings, e.g.
   critical · prod-cluster · storage container near full · 93.0% used >= 90.0% threshold
2. storage_container_list → confirm which container it is and what its
   maxCapacityBytes / logicalUsageBytes actually are
3. recovery_point_list / snapshot_list <vm_ext_id> → the usual culprit is
   snapshot sprawl in that container
4. snapshot_delete <…> --dry-run to preview, then re-run to reclaim space —
   optionally set NUTANIX_AUDIT_APPROVED_BY first to annotate who authorized
   it; each deletion is audited and records an undo descriptor
5. Re-run diagnose cluster-health → the finding should drop below the 80%
   warning threshold; the before/after percentages are your evidence

Triage a cluster alert (RCA)

1. nutanix-aiops diagnose alert-triage (or alert_list) → per-severity counts and the oldest unresolved alert, so you know which extId to open first
2. analyze_alert <alert_ext_id> → probable cause + suggested actions, built by correlating the alert with related event_list records
3. Confirm blast radius with cluster_health / cluster_utilization, then alert_acknowledge (or alert_resolve once fixed)

Safely delete a VM (high-risk, audited)

1. vm_get <vm_ext_id> → confirm it's the right VM (and see its ETag)
2. nutanix-aiops vm delete <vm_ext_id> --dry-run → preview the exact DELETE call
3. Optionally annotate who/why: export NUTANIX_AUDIT_APPROVED_BY=… NUTANIX_AUDIT_RATIONALE=…
4. Re-run without --dry-run (double-confirm at the CLI); the call is audited with
   tier and any approver/rationale supplied

Snapshot sprawl cleanup

1. recovery_point_list (or per-VM snapshot_list <vm_ext_id>) → find stale / redundant snapshots
2. snapshot_delete <…> --dry-run on each candidate → preview
3. Re-run without dry-run (HIGH risk, double-confirm at the CLI) to reclaim space; each deletion is audited

Capacity runway

1. cluster_utilization <cluster_ext_id> → current CPU / memory / storage / IOPS
2. capacity_runway → days-to-full forecast per resource; use it to schedule an lcm expansion or storage add before you hit the wall

Migrate a VM to another host (reversible)

1. host_list → pick the destination host extId
2. nutanix-aiops vm migrate <vm_ext_id> <target_host_ext_id> --dry-run → preview
3. Re-run without dry-run (HIGH, double-confirm); the prior host is captured as an undo descriptor so a regression can be reversed

Governance & Safety

The skill delivers reads and writes and records them; it does not decide whether a write is
permitted. That is your agent's judgement, or the permission of the account you connect it with
(connect with a Prism Central account holding only a read-only (Viewer) role — writes then fail
at the server). There is no read-only switch, policy file, or approval gate.

• Audit is the guarantee, and it is not bypassable. Every operation — MCP and CLI alike — is logged to ~/.nutanix-aiops/audit.db (relocatable via NUTANIX_AIOPS_HOME): params, result, status, duration, and the risk tier. The CLI writes the same row the MCP path does.
• NUTANIX_AUDIT_APPROVED_BY / NUTANIX_AUDIT_RATIONALE are optional annotations recorded on the audit row (who/why); they are never required and never block.
• Runaway guard — a safety backstop, not authorization: the same call looped in a tight window trips a circuit breaker. Disable with NUTANIX_RUNAWAY_MAX=0.
• Every mutation auto-handles ETag / If-Match; every list auto-paginates.
• Destructive writes support dry_run / --dry-run and, at the CLI, double confirmation.
• Reversible writes record an undo descriptor (vm_update → prior CPU/memory, vm_migrate → prior host).

References

• references/capabilities.md — full 51-tool + API-path reference
• references/cli-reference.md — CLI command reference
• references/setup-guide.md — onboarding, credentials, REST-RBAC, CE self-test
• references/agent-guardrails.md — which guardrails the harness enforces for you, and a ready-to-paste system prompt for smaller / local models
  </content>