Agent: install, update and options
The agent is a single program on each machine. It opens one outbound TLS connection to wecommander.app, never listens on a port, and enforces the device's limits itself: allowed folders, read-only and no-command switches, secret masking and signed approvals for high-risk actions.
How it runs on each system
| System | Installed by | Runs as | Default allowed folder |
|---|---|---|---|
| Windows desktop | The WECOMMANDER app (agent inside) | Your Windows user, as a hidden child of the app | C:\Users\<you>\WECOMMANDER |
| Windows server | install/agent.ps1 (setup) | The signed-in user, through a per-user scheduled task | Chosen during setup |
| macOS | The WECOMMANDER app (agent inside) | Your macOS user | ~/WECOMMANDER |
| Linux server | install.sh (Debian package) | The unprivileged wecommander system user (systemd service webrising-commender) | /var/lib/webrising-commender/workspace |
Pairing uses a one-time request approved in your browser; the device credential is stored on the device and never printed. Re-running an installer keeps the existing device identity.
Linux options
# Change the allowed folder or switches on an installed server
sudo wecommander-setup --allowed-root /srv/app
sudo wecommander-setup --read-only # refuse all file writes
sudo wecommander-setup --no-commands # refuse commands, tasks and terminalsChoose a narrow folder the wecommander user can actually access; do not start with /, /root, /etc or a whole data disk. Commands run with the wecommander user's Unix rights, so grant it access only to what the AI should use.
Device switches the gateway cannot change
| Switch | Refuses |
|---|---|
| Commands off (--no-commands) | process_exec, process_kill, task_start, terminal_start, terminal_input |
| File writes off (--read-only) | fs_write, fs_append, fs_mkdir, fs_delete, fs_move, fs_replace, fs_undo, xlsx_write_cells |
| Root not allowed (default) | any action the gateway sends to run as root |
| Output redaction (on by default) | turn off with capabilities --redact-output off |
| System logs (on) | turn off journal reading with capabilities --system-logs off |
These live only in the agent's local config. Even a valid human approval cannot override them. To change them later run webrising-commender --config <path> capabilities --commands on|off --file-writes on|off and restart the service. policy_info shows what is on.
Allowing root (Linux only)
Running as root needs two keys. The first is the machine owner's, on the machine:
sudo bash /opt/wecommander/libexec/install-linux.sh --allow-root # allow
sudo bash /opt/wecommander/libexec/install-linux.sh --disallow-root # undoThis adds a sudoers rule for the agent's service user, removes the systemd hardening that blocks sudo and records the ceiling in the agent's config. The second key is an admin turning “run as root” on for that device in the dashboard, with their password. Root is not available on Windows or macOS agents. See AI permissions.
Check and update
sudo systemctl status webrising-commender --no-pager
sudo journalctl -u webrising-commender -n 30 --no-pager- Linux: run the install command again to update; the device stays the same.
- Windows and macOS: the agent comes inside the app and is updated with it (Update in the Windows app, Update app on the Mac).
- The dashboard shows each device's agent version under Devices.
- Download and install: Install the WECOMMANDER app on Windows (installer, PowerShell or Microsoft Store) and macOS, the agent on Linux and Windows servers, and the Management Center on Linux desktops.
- AI permissions: trust levels, allowed commands, root and the safety net: Set per device and per SSH server what AI may do without asking: strict, standard or full trust (time-limited or not), allowed command patterns, run as root with two keys, the safety net and autonomous mode.
- Troubleshooting: Fixes for common problems: agent offline, host key changed, approval not arriving, SSH server unreachable for AI, root not available, commands refused, updates and plan limits.