What it does
VigilOps runs as one container on a Docker server and reads the Docker socket to see what is running. Every 60 seconds it checks containers, disk, memory, nginx and health URLs. It restarts containers that exited and prunes images when disk usage passes a set threshold. A container that keeps crashing is escalated instead of restarted again. Backups dump a database, compress it with gzip and copy it off the server. They can run hourly, daily or weekly. Deploys pull the new image, restart the service, wait for the health check and roll back if it fails. A deploy can be triggered from Telegram, the command-line interface or a webhook in continuous integration. Problems that no rule covers go to an AI provider the user chooses, or to no AI at all. Every action is written to an audit log on the server.
Capabilities
- A restore test restores a backup into a throwaway container with no network, counts tables and rows against the live database, and removes the container afterwards.
- Restores can target the same container or a different one, and a safety backup is taken first.
- Every command VigilOps runs on its own must match a short allowlist, and anything else is blocked and written to the audit log.
- Risky actions wait for approval on Telegram and are denied after 10 minutes.
- Thirteen MCP tools let Claude Code, Cursor and VS Code read the state of the server and act inside the same allowlist.
When to use it
- Watching containers and backups on the one Docker server you already have.
- Running a restore test of a database backup.
- Reaching a server from Telegram, or from Cursor and Claude Code.
