Approvals: web, e-mail and desktop notifications
When an AI calls a high-risk tool that the device's AI permissions do not allow on their own, WECOMMANDER does not run it. It stores the exact requested action as an approval request and tells the AI to try again once a person decides.
The flow
- The AI calls, for example, process_exec with a command. It gets approval_required and an approval_request_id.
- Admins and owners are notified. The request shows the device or server, the exact command, file, process or container, and whether it would run as root.
- A person approves or denies it.
- The AI repeats the same call. The gateway uses the approval once, signs it for that device, and the agent checks the signature before running.
What one approval allows
- Exactly the action shown, on that device, once. Changing one byte of the command, path, PID or terminal input needs a new approval, and a replay is refused.
- Requests for agent devices expire after 10 minutes; requests for SSH servers after 5 minutes.
- If the device did not receive an approved action (it went offline), the approval is kept for the identical retry.
- Only admins and owners can approve, and only from a browser session or a connected Management Center. Requests for an SSH server shared through someone's computer are decided only by that person.
Where requests arrive
| Place | How |
|---|---|
| Web dashboard | AI → Approvals lists pending requests with Approve and Deny. |
| Admins and owners get one e-mail per new request with a link to Approvals. The e-mail itself grants nothing; approving still needs your browser session. | |
| Windows | The Management Center shows each request as a Windows notification with Approve and Deny buttons, and can open its panel automatically. A test notification checks that Windows lets them through. |
| macOS | macOS notifications from the WECOMMANDER app, and the app's panel. |
| Linux desktop | The Linux Management Center's Approvals tab and notifications. |
Connecting a Management Center
The desktop app signs in through the browser: you approve “Connect this Management Center” with an admin or owner account, and the app receives its own token; your password never enters it. That token can only list and decide approvals, read devices, manage autonomous mode and scheduled tasks; it cannot run device tools. Revoke a Management Center under Approvals, or sign out in the app.
Fewer approvals, safely
- Allow routine commands (standard trust) instead of approving df -h every time.
- Use full trust for a set time on a device you are actively working on.
- Autonomous mode for several devices at once, with automatic expiry.
- 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.
- Audit log and undo snapshots: What the WECOMMANDER audit log records for every AI tool call, approval and permission change, and how undo snapshots let you restore a file the AI changed.
- 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.