proxmox-ops-mcp
access level: configurable✔ Source-reviewed Oct 2026
We read this project's source code and tool definitions (Oct 2026). This is a code review, not a live test against real systems. Re-check after upgrades.
| Repository | https://github.com/derdennis1012/proxmox-ops-mcp |
|---|---|
| Category | Proxmox |
| Systems touched | Proxmox VE clusters: nodes, guests, Ceph, ZFS/SMART, backups, HA |
| Access (as described) | configurable |
| License / price | MIT / free |
| GitHub stars | 0 |
| Read tools | cluster_health, ceph_health, compare_nodes, backup_audit, recent_task_failures, storage_health, node_maintenance_status, recommend_placement, plan_node_evacuation, list_nodes, list_guests, get_guest (secrets redacted), get_task_status |
| Write / delete / exec tools | guest_power (start/shutdown/reboot; stop needs admin tier), create_snapshot (operator tier), migrate_guest (admin tier); no delete tools |
| Read-only mode | Yes: PROXMOX_ACCESS_TIER defaults to read_only and higher-tier tools are not registered; writes are two-phase (preview, then one-time 5-minute token bound to exact params); optional pool/VMID limits, hourly caps and audit log |
| Safety notes | Start with a PVEAuditor token (README shows the pveum commands). The README is honest that the confirm token proves "what runs is what was previewed", not that a human approved, because the model sees the token, so keep client approval on for write tools. |
| Verified | yes (source review 2026-10-31) |
| Notes | Brand new (created 27 Oct 2026), v0.1, one author, 0 stars. Ops-first: backup_audit, plan_node_evacuation, storage_health instead of raw endpoints. On PyPI and GHCR. Added 2026-10-31. |
📬 SysAdmin Agents
The SysAdmin Agents newsletter: new MCP servers, agent skills and safety notes for sysadmins. One email a week.