Native SSH Remote Projects and Remote Agent Execution

Agentsroom currently works well with local projects, but working with projects located on a remote development machine requires a fairly complicated workaround.

A possible workaround is to mount the remote project directory through SSH/SFTP so that it appears to Agentsroom as a local filesystem. The mounted directory can then be opened as a normal project.

While this technically works, it has several disadvantages:

• It requires external filesystem mounting and additional system configuration.
• The mount has to remain available or be recreated automatically after restarting the local machine.
• Cross-platform support for such setups can be inconsistent and may require different solutions depending on the operating system.
• Opening the files remotely does not automatically mean that agents execute on the remote machine.
• Additional per-project configuration may be required to make agents execute remotely.
• Local project handling and remote execution can become disconnected from each other.
• Some settings may not behave consistently when the project is only exposed through a filesystem mount.

A much cleaner solution would be native support for remote projects over SSH, similar to remote-development workflows available in modern IDEs.

Ideally, Agentroom could connect directly to a remote machine through SSH and allow the user to select a project directory there.

Possible workflow:

1. Add an SSH host to Agentsroom.
2. Configure authentication using an SSH key, SSH agent, or another supported SSH authentication method.
3. Connect to the remote host.
4. Select a directory on the remote host as a project.
5. Agentsroom displays the remote project like a normal project in the UI.
6. Agent processes, commands, terminals, and project-related operations execute on the remote host by default.
7. Project-specific Agentsroom settings continue to work normally.
8. Reconnecting