One console over Proxmox VE clusters and XCP-ng pools
PegaProx is a self-hosted web console that brings several Proxmox VE clusters, and XCP-ng pools, into one interface. Written in Python (Flask), it runs on its own machine and reaches each cluster through the Proxmox API, with SSH on top for some features. The project's website describes PegaProx as a layer on top of Proxmox VE "without replacing it", one that adds cross-cluster migration, load balancing, and cost and compliance reports.
Proxmox Datacenter Manager is the console Proxmox publishes for its own products. ProxCenter is a third-party console, like PegaProx.
The comparison of multi-cluster consoles for Proxmox weighs PegaProx against the vendor's console and the other third-party ones.
Customer isolation and migration off ESXi
- XCP-ng next to Proxmox: on an XCP-ng 8.2 or later pool, reached over XAPI, PegaProx powers VMs on and off, opens their console, manages disks and network interfaces, and puts a host into maintenance once its VMs are evacuated.
- Moving VMs between hypervisors: ESXi to Proxmox VE or XCP-ng, and Proxmox VE to XCP-ng or back, with disk format conversion. Snapshots stay behind.
- Fleet-wide operations: load balancing across nodes and across clusters, rolling node updates that evacuate each node first, snapshot policies with retention, and since 1.3.0, bulk actions on selected guests.
- Access split by customer: custom roles, per-VM ACLs, clusters isolated per customer, LDAP or OIDC sign-in, FIDO2 keys. The Client Portal plugin gives a hosting provider's customers the console and snapshots of their own VMs.
- Cost and compliance: a monthly cost per VM computed from real metrics, and a read-only dashboard mapped to BSI Grundschutz, ISO 27001, NIS2 and SOC 2, exportable as PDF.
- A console with a standby: since 1.3.0, up to four PegaProx instances can form a group around one leader, and an admin promotes a standby by hand when the leader is gone.
Four features need root, and the portal can't be white-labeled
The root account is the recommended login. The docs "strongly recommend" root@pam with each cluster's root password, which PegaProx keeps encrypted in its database. With an API token or a limited user, four features stop working: rolling updates, SMBIOS auto-config, two-node HA and the node shell.
XCP-ng is a tech preview. The README files the integration under "Tech Preview", and the docs warn that some features may be limited or change.
Cross-cluster migration rests on an experimental Proxmox feature. PegaProx calls the Proxmox VE remote_migrate API (pegaprox/core/manager.py), the one behind qm remote-migrate, which the Proxmox VE documentation still flags as "EXPERIMENTAL feature!". Proxmox Datacenter Manager goes through the same API.
The AGPL comes with an attribution term. The NOTICE file invokes section 7(b) of the AGPL-3.0: the "Powered by PegaProx" line in the client portal and the copyright line in the main interface must stay visible on every instance users can reach. A hosting provider therefore cannot white-label the portal. Every feature is free; commercial support comes from independent companies listed as project sponsors.
Python and SQLite, no database server. PegaProx runs on Python 3.10 to 3.13 on Linux, or as a Docker image, and keeps its configuration in one SQLite file, fully encrypted with SQLCipher on Linux x86_64 only. The docs recommend 4 cores, 8 GB of RAM and 20 GB of disk for production; up to five clusters, they count 1 core and 1 GB of RAM. The README's install script pulls the main branch, not a tagged release; for production, the README points to an APT repository run by a contributor, gyptazy.
The website lists SAML; the README does not. The comparison table on pegaprox.com ticks SAML, but the README names only LDAP and OIDC, and no SAML library appears in requirements.txt.
57 releases since January 2026. The repository was created on 18 January 2026. Version 1.0 ended the beta on 1 August 2026, and six releases have followed, three of them minor, up to 1.3.0 on 7 October 2026. Two developers account for 1,095 of the 1,296 commits, and the README describes a three-person team funded by sponsorship. The 1.3.0 notes list 38 behaviour changes and a fix for a silent data loss in the ESXi "Live Clone" migration. They also leave a known issue open: the "Snapshot-Iterative" ESXi migration can report success while data is missing on the target.