At a glance
| Point | VPS + SSH + tmux | OpenClowd |
|---|---|---|
| Compute owner | You choose and operate the machine. | You still choose and operate the machine for the planned self-hosted release. |
| Session view | Terminal names, panes, scripts, and memory. | A planned workspace view for sessions, status, history, files, and diffs. |
| Remote access | SSH, VPN, reverse proxy, or tunnel that you configure. | A documented HTTPS and authentication path is a launch requirement, while host security remains yours. |
| Mobile use | Usually an SSH client and a lot of pinching and zooming. | A mobile-friendly web companion for checking, reattaching, and sending short input. |
| Process survival | tmux can survive disconnects, but not a dead or rebooted host. | The same honest boundary: reattach while the host is alive, with no promise that live processes survive a reboot. |
| Availability today | Available now with mature tools. | In development. The verified public installer has not shipped. |
Credit where it is due
You already made the important ownership decision
If your agents run on a VPS or server you control, the repository, terminal processes, and host credentials are not trapped inside an agent vendor's product. That is the right foundation. OpenClowd is not an argument for giving it up.
The product bet is narrower: owning the compute should not require remembering which city-named tmux session contains the one agent waiting for an answer.
Keep the machine. Keep the accounts. Replace the improvised control surface.
The product layer
OpenClowd is the missing control plane, not the missing server
The planned self-hosted product organizes real repositories and terminal agents on an always-on host. Its workspace model brings sessions, lightweight attention signals, file and diff context, Git awareness, and app previews into one browser surface.
It will not provision your VPS, patch the operating system, install every agent, or become a secrets manager. Self-hosting still means operating the host. The difference is that the daily agent loop becomes a product instead of a personal collection of shell habits.
The decision
Choose based on whether the plumbing is part of the fun
Both paths preserve control. The difference is how much workflow infrastructure you want to own yourself.
Choose DIY if you need something today, enjoy owning every layer, and are comfortable composing SSH, tmux, Git, networking, and mobile workarounds yourself.
Follow OpenClowd if you want that same developer-controlled foundation with a purpose-built, browser-based agent workflow once the public release is ready.
Questions, answered
What people usually ask next
01Does self-hosted OpenClowd control my VPS?
No. You supply and maintain the machine, operating system, network exposure, repositories, and third-party agent credentials. OpenClowd is the control plane that runs on that host.
02Will sessions survive a server reboot?
The v1 promise is reattachment, status, and history while the host is alive. OpenClowd does not promise that a live agent process survives a reboot or crash.
03Can I install OpenClowd now?
Not yet. A verified one-command installer is planned for the public release, but the site does not publish an install command before that path is tested and documented.