Security at ServerGuard
ServerGuard lets you remove your company's source code from a developer's machine the moment their access ends. This page explains exactly how that works, what happens when a machine is offline, and what no endpoint product can prevent.
How a revoke works
The ServerGuard agent runs on each enrolled developer machine and checks in with the ServerGuard control plane every few seconds over TLS. Each device authenticates with its own token, issued at enrollment. When an admin revokes a license or device — or an HR integration reports that someone has left — the next check-in tells the agent to revoke. It then:
- Pushes any uncommitted or unpushed work to a quarantine branch on your own git remote, so nothing your company owns is lost.
- Deletes the protected working copy from the machine, file by file.
- Destroys the device-bound encryption key (macOS Keychain, Windows DPAPI, or the Linux Secret Service), so local encrypted copies can no longer be decrypted.
- Leaves a revoke marker so the agent cannot silently re-enroll.
The result of each step is reported back to your dashboard.
When a device is offline
Every successful check-in grants the device a lease. If the device stops checking in, it keeps working until your organization's Offline Timeout runs out (configurable from 5 minutes to 24 hours, or Never, on the Security Policies page). The clock survives restarts, so rebooting does not reset it.
When the timeout runs out, the agent follows your policy:
- If you made Offline timeout exceeded a Repository Lock Condition (the High preset does), the agent runs the full revoke described above.
- Otherwise, the device keeps working and an offline-timeout violation appears on your dashboard as soon as it reconnects.
Our outages never wipe your developers. If a device can reach ServerGuard but our service returns an error, the agent knows the problem is on our side and holds instead of revoking — for up to 72 hours. Network failures that look the same as deliberate blocking (DNS, refused connections, timeouts) are still treated as offline.
Data we handle
- Your source code is never uploaded to ServerGuard. Encryption and deletion happen on the developer's machine.
- We store account, license and device data, plus security events (for example file paths involved in a copy or upload detection) so you can investigate them.
- Dashboard accounts support TOTP two-factor authentication with one-time recovery codes. Passwords are hashed with Argon2.
- Everything between agents, browsers and ServerGuard is encrypted with TLS.
- Backups are encrypted before they leave our servers.
What endpoint protection cannot do
We would rather you hear this from us. ServerGuard reduces the risk of source code leaving with a developer. It cannot eliminate it. A revoke cannot undo:
- copies already made to USB drives, other machines, or personal cloud storage
- code already pushed to forks or other remotes
- screenshots, photos of the screen, or what someone remembers
- machines where the agent was never installed, such as personal laptops
ServerGuard detects many of these attempts — copies, archives, cloud-sync folders, clipboard use, decoy files, and clones from unexpected IPs through the GitHub audit log — so you can act quickly. Pair it with access controls on your git host, such as IP allowlists and SSO.
Reporting a vulnerability
If you believe you have found a security issue in ServerGuard, please contact us through your account dashboard or reply to any ServerGuard email. Please do not publish details until we have had a chance to fix it.