Governed AI operations tooling for infrastructure — audit, budget, undo, and graduated approval built into every tool.
AI agents are great at infrastructure ops, right up until something goes wrong in production. AIops-tools is a family of MCP servers + CLIs that wrap real infrastructure platforms with a governance harness, so an agent's actions are:
- Audited — every operation logged to a local SQLite trail (who / what / why / when), secret-redacted.
- Bounded — per-process token/call budget + a runaway-loop breaker (no more "one operation, 26k tokens").
- Reversible — write operations record an inverse undo token wherever a clean inverse exists.
- Graduated — risk tiers gate writes by environment/tag; the highest tiers require a recorded approver.
- Safe by default — destructive ops need double confirmation +
--dry-run; all API text is sanitized.
Every tool is self-contained (the harness is bundled — no shared runtime dependency) and ships on PyPI, the MCP Registry, and ClawHub.
| Tool | Platform | Install | MCP tools |
|---|---|---|---|
| proxmox-aiops | Proxmox VE — VMs, LXC containers, snapshots, cluster, storage | pip install proxmox-aiops |
23 |
| veeam-aiops | Veeam Backup & Replication — jobs, restore, repositories, sessions | pip install veeam-aiops |
12 |
| k8s-aiops | Kubernetes — k3s / EKS / GKE / AKS (pods, deployments, nodes) | pip install k8s-aiops |
15 |
| network-aiops | Network devices via NAPALM — Cisco IOS/NX-OS/IOS-XR, Arista EOS, Juniper Junos + NetBox | pip install network-aiops |
13 |
Each tool exposes both a CLI and an MCP server (<tool> mcp, stdio transport).
Point any MCP client (Claude, etc.) at it. Configuration and the audit/undo/policy
store live under ~/.<tool>/ (relocatable via <TOOL>_AIOPS_HOME).
pip install proxmox-aiops # or pipx / uv tool install
proxmox-aiops doctor # verify connectivity
proxmox-aiops mcp # run as an MCP serverMIT licensed. Community-maintained. Not affiliated with, endorsed by, or sponsored by the vendors of the platforms these tools operate (Proxmox Server Solutions GmbH, Veeam Software, the CNCF / Kubernetes project, etc.); all trademarks belong to their respective owners.
