X2Go has one trick that most remote desktop tools still cannot match: you close the client, go home, open it on another laptop, and the session is exactly where you left it.
X2Go has one trick that most remote desktop tools still cannot match: you close the client, go home, open it on another laptop, and the session is exactly where you left it. Terminals still running, the half-written email still open, the build log still scrolling. For a Linux workstation on the same network, or a server with a public IP and SSH open, it is hard to beat. The trouble starts the moment the machine you need sits behind a home router, a carrier-grade NAT or an office firewall that nobody will open port 22 on. X2Go has no answer for that, because it was never designed to have one.
This guide covers what X2Go actually is under the hood, the specific situations where it stops working, and the alternatives worth considering for each of them. If you are coming from the other common Linux client, our Remmina alternative guide covers similar ground from the client side.
What X2Go is, in one paragraph
X2Go is a GPL-licensed remote desktop system built on the NX 3 compression libraries originally written by NoMachine. The server component, x2goserver, runs only on Linux (and in practice mostly on Debian, Ubuntu, Fedora and openSUSE). Clients exist for Linux, Windows and macOS. Every session is tunnelled through SSH: the client logs in over port 22 with a password or key, starts a fresh X session for that user on the server, and streams it back. Audio goes through PulseAudio, shared folders are mounted with SSHFS, and printing can be forwarded to the client. There is no account system, no broker, no relay and no cloud. The security model is simply SSH.
Where X2Go is genuinely good
Before looking at replacements it is worth being clear about what you would give up, because some of it is unusual.
- Suspend and resume. A session belongs to the server, not the connection. Drop the Wi-Fi, close the lid, come back two days later from a different client, and resume. This is the feature people cite when they refuse to switch.
- Separate virtual desktops per user. Ten people can each have their own desktop on one server at once, none of them touching the physical display. That makes X2Go a cheap terminal server for a small team.
- Good performance on slow links. NX 3 caches and compresses X11 drawing operations rather than sending raw pixels, so a lightweight desktop like XFCE or MATE stays usable at 1-2 Mbit/s.
- Published applications. You can run a single remote program in its own window instead of a whole desktop.
- Nothing to pay and nothing to register. If you already run SSH, the infrastructure cost is close to zero.
If all your machines are on one LAN, or each has a public address with SSH reachable, and everyone uses XFCE, X2Go may be the right tool and you can stop reading here.
Where it stops working
1. The machine is behind NAT and you cannot forward a port
X2Go needs a direct SSH connection from the client to the server. A home PC behind a consumer router, a box on a mobile hotspot, or anything behind a provider's CGNAT has no inbound address you can reach. The usual workarounds are port forwarding (often impossible, see our article on remote desktop behind CGNAT), a VPN such as WireGuard or Tailscale, or a reverse SSH tunnel to a VPS you keep running. Each works, and each is one more moving part you now maintain. Our guide to remote desktop without port forwarding walks through why NAT blocks inbound connections in the first place.
2. Port 22 is exposed to the internet
The flip side: if you do forward port 22, you now run an internet-facing SSH server. Within hours of going live it will see thousands of password attempts a day. That is manageable with key-only authentication, fail2ban and a non-standard port, but it is a security job, not a remote desktop setup. Corporate networks usually refuse it outright, and many block outbound SSH as well; see remote desktop through a corporate firewall.
3. Modern desktops: GNOME, KDE Plasma and Wayland
X2Go starts X11 sessions. GNOME 3 and newer KDE Plasma releases depend on compositing that the NX 3 agent handles poorly, so users get black windows, broken animations or sessions that refuse to start. The X2Go wiki itself recommends XFCE, MATE, LXDE or Cinnamon in fallback mode. And it cannot attach to a Wayland session at all, which is now the default on Ubuntu, Fedora and Debian with GNOME. If your users expect their normal desktop, that is a real limit; the Wayland remote desktop support article explains what does and does not work today.
4. Seeing the physical screen
X2Go's strength, a separate virtual session, is a weakness for support. If a colleague calls with an error dialog on their monitor, you need to see that monitor, not a fresh desktop. X2Go has a desktop sharing mode, but it requires the companion tool on the remote side, works only for X11 displays, and is rarely set up in practice.
5. Linux only on the server side
You cannot use X2Go to reach a Windows PC or a Mac. Mixed fleets need a second tool anyway, and the macOS client additionally depends on XQuartz being installed.
6. Slow release cadence
X2Go is maintained by a small volunteer team. Releases come years apart and Windows client builds sometimes lag behind current OS versions. That matters if you need fixes on a schedule.
The alternatives, matched to the problem
| Tool | Keeps sessions after disconnect | Works behind NAT without port forwarding | Server OS | Price |
|---|---|---|---|---|
| X2Go | Yes | No (needs SSH reachable) | Linux | Free |
| NoMachine | Yes | Only with paid Cloud Server or port forwarding | Linux, Windows, macOS | Free for personal use; Enterprise licences per server |
| xrdp | Yes (per-user Xorg sessions) | No (needs port 3389 or a VPN) | Linux | Free |
| TigerVNC over SSH | With vncserver, yes | No | Linux, Windows | Free |
| Self-hosted RustDesk | No (attaches to the live desktop) | Yes, through your own relay on a public server | Linux, Windows, macOS | Free software; you pay for and run the server |
| Tenvo | No (attaches to the live desktop) | Yes, through our managed relay | Linux, Windows, macOS | Free $0, Lite $2.99/mo, Pro $7.99/mo |
NoMachine: the closest relative
NoMachine wrote the NX protocol X2Go is built on, and its current NX 4+ engine handles modern desktops much better than NX 3. It supports suspend and resume, virtual desktops on its Workstation and Terminal Server editions, and runs on all three major operating systems. The free edition is for personal use and shares the physical display only; virtual multi-user sessions need a paid server licence. Reaching a machine behind NAT still means port forwarding or NoMachine's own paid cloud server. Our NoMachine on Linux alternatives article goes into the editions in detail.
xrdp: if your clients are Windows
xrdp exposes a Linux machine over the RDP protocol, so Windows users connect with the built-in Remote Desktop Connection client. It gives each user their own Xorg session and keeps it when they disconnect, which covers the X2Go use case on a LAN. GNOME needs some tuning (a separate session type and polkit rules), and over the internet it has the same reachability problem, with the extra risk that port 3389 attracts even more scanning than SSH. The trade-offs between protocols are covered in VNC vs RDP for modern remote access.
VNC tunnelled through SSH
TigerVNC's vncserver can run persistent virtual desktops, and wrapping it in an SSH tunnel solves encryption. It is the most manual option: you manage display numbers, passwords and tunnels yourself, and it is heavier on bandwidth than NX. Useful as a fallback, rarely as the main tool.
Self-hosted RustDesk: when a written rule requires your own server
RustDesk solves the NAT problem by having both ends connect outward to a rendezvous server, then attempting a direct hole-punched connection and falling back to a relay. You can run that server yourself on a VPS with a public IP. This is the right choice when a compliance requirement bars third-party infrastructure or the network is isolated. Outside that case, count the real cost first: a server to patch, keys to back up, a relay to keep running at 2 a.m. Our article on the hidden costs of self-hosting adds them up, and Linux remote desktop server covers the setup.
Tenvo: reach the box without opening anything
Tenvo is built on RustDesk and runs the connection infrastructure for you. You install the client on the Linux machine, sign in, and it appears in your device list. Both ends make outbound connections only, so there is no port to forward, no SSH server to expose and no VPN to configure; it works from behind CGNAT and from a phone hotspot. When a direct peer-to-peer path is possible, the session goes straight between the two devices. When it is not, traffic falls back to our managed relay, two servers in Frankfurt, Germany.
Sessions use TLS with a per-device certificate. Be aware of what that means on the relay path: a direct connection is protected end to end between your two machines, while a relayed session terminates TLS at the relay we operate. If that distinction matters for your threat model, remote desktop encryption explained lays out how each design works. Clients exist for Linux, Windows and macOS, with a browser client in public beta, and the client is licensed under AGPL-3.0.
What you give up compared to X2Go: Tenvo shows the desktop that is actually running on the machine, so it does not create a separate virtual session per user, and it does not have X2Go's suspend-and-resume of an independent session. For remote support, for reaching your own workstation, or for a mixed fleet of Linux, Windows and Mac machines, that is usually what you want. For a ten-user terminal server on one LAN, it is not.
Pricing, in plain numbers
- Free, $0: non-commercial use, up to 3 devices, 1 session at a time, 2 hours a day in total, 5 GB a month of relay traffic.
- Lite, $2.99 a month: up to 50 devices, no daily time cap.
- Pro, $7.99 a month: for heavier professional use.
For comparison, keeping X2Go reachable from outside typically means a small VPS for a reverse tunnel or a commercial VPN, which costs about the same as Lite before you count the time spent maintaining it. For a side-by-side with the project Tenvo is built on, see RustDesk vs Tenvo.
How to choose
- All machines on one LAN, lightweight desktops, users want persistent sessions: keep X2Go, or move to xrdp if your users are on Windows.
- You need modern GNOME or KDE with persistent sessions: NoMachine is the closest drop-in, with a paid licence for multi-user virtual desktops.
- A written requirement says the server must be yours: self-hosted RustDesk on your own public server.
- The machine is behind NAT, CGNAT or a firewall you do not control, or the fleet mixes Linux with Windows and Mac: a managed relay such as Tenvo, with no ports and no tunnels to maintain.
Plenty of teams end up running two tools: X2Go or xrdp for the in-office terminal server, and something relay-based for everything outside the building. That is a reasonable split. What is not reasonable is maintaining a reverse SSH tunnel on a forgotten VPS just so X2Go can pretend to be internet-reachable. If you are setting up a headless box, Linux headless remote desktop covers the extra steps a machine without a monitor needs.
Getting started
If the machine you keep failing to reach with X2Go sits behind somebody else's router, install Tenvo on it and on your laptop, sign in on both, and connect by device ID. The step-by-step Tenvo tutorial walks through the first session, and if a firewall still blocks you, remote desktop blocked by a firewall lists what to check. X2Go can keep the LAN.
