ProxmoxMCP-Plus v0.6.3

Changes

• Add get_node_time for node clock and timezone inspection, get_node_storage for capacity on the selected node, and get_node_diagnostics for CPU, memory, storage and observed RRD I/O metrics. Missing optional metrics produce partial results instead of a false completeness claim.
• Add get_cluster_join_info, exposing documented public cluster metadata and certificate fingerprints while omitting credentials and unrecognized extensions.
• Add read-only preflight_container_migration and guarded migrate_container. Both use fresh checks for node availability, guest state, rootfs storage and capacity, mounts/devices, HA membership and matching backup inventory. Missing or malformed evidence blocks readiness.
• Submit migration with LXC-supported API parameters and explicit migration/downtime confirmations. Apply the selected target's configured high-risk policy, return the submitted task UPID and optional local job ID, and never automatically retry migration. Ambiguous transport or submission responses return SUBMISSION_UNCERTAIN.
• Keep reachable-node backup results when other cluster members are offline, and report an incomplete inventory instead of a false empty result.
• Align the new tools across the manifest, canonical catalog, read-only policy and operator documentation. Preserve contributor attribution from #167–#172 through the reviewed integration in #174; the offline-backup fix is #173.

Validation

The included runtime changes passed GitHub CI on Python 3.11 and 3.12 with PostgreSQL: 1,043 tests passed and 2 optional integration tests skipped on each version. All 61 runtime modules have zero uncovered statements and all 9 critical modules have zero uncovered branches. Windows contracts, Ruff, Mypy, package checks, dependency audits, secret scanning and CodeQL passed. The release metadata update follows the same required CI gates before publication.

No live Proxmox migration was performed. Migration assessment is advisory: it does not reserve capacity or verify backup freshness/restorability, application health or every Proxmox migration constraint. Complex mounts, unused volumes, custom devices and HA-managed guests require separate operator handling.

Upgrade

Upgrade to proxmox-mcp-plus==0.6.3 or ghcr.io/rekklesna/proxmoxmcp-plus:0.6.3, restart workers, and reconnect MCP clients to refresh tool discovery. Preserve existing SQLite state and legacy OAuth metadata; no database migration is introduced.

The default high-risk operation list now includes migrate_container. If you supply a custom command_policy.high_risk_operations list, add this tool explicitly to apply your configured approval policy. confirm_migration=true is required for every migration; restart=true also requires confirm_downtime=true. Default audit-only mode remains available, and read-only targets reject migration submission.

After submission, poll the job/task and inspect the destination guest. If SUBMISSION_UNCERTAIN is returned, inspect Proxmox tasks and guest location before considering another call.