Server monitoring · Agent security

A monitoring agent that can only read

Installing an agent means trusting it with your server. Ours is built so that the worst it can do is report numbers: it cannot be told to run, restart or change anything.

The short version

  • One Python file, standard library only. You can read all of it before you install it.
  • It opens no port and accepts no commands. It only makes outgoing HTTPS requests to ismysiteup.
  • It runs as its own user with no shell and no sudo, inside a locked-down systemd service.
  • The installer verifies the file's SHA-256 before installing it.
  • It is capped at 20% of one CPU core and 128 MB of memory.

Read the source

The agent and its installer are served as plain text. Open them in a browser or download them and read them before running anything.

The two files
https://ismysiteup.net/imsu-agent.py
https://ismysiteup.net/install-imsu-agent.sh

See exactly what it would send

Run the agent with --check and it prints the full report it would send, then exits without sending anything.

On the server
sudo -u imsu-agent python3 /opt/imsu-agent/imsu-agent.py --check

What it reads

  • CPU, memory, disk and network counters from /proc and the filesystem.
  • The busiest processes: user, CPU, memory and command line. Passwords, tokens and keys in command lines are masked on the server before anything is sent.
  • Network counters that show a flood (connection requests, drops, packets). Numbers only, no addresses.
  • Security signs: which ports are open, the two SSH login settings, the names of accounts that can log in, counts of outgoing mail and SSH connections, failed services. Names, ports and counts only: no file contents, no addresses.
  • Optionally the Docker container list, and NVIDIA GPU numbers when nvidia-smi is installed.

What we can tell the agent

Two things, and nothing else. How often to report, and that you removed this server from your account. On removal the agent stops, and unless you installed it with --keep-files it uninstalls itself. That removal is proven with the server's own secret and runs a fixed local script; nothing in it comes from the network.

How reports are protected

  • Each server has its own secret. Every report is signed with it (HMAC-SHA256), with a timestamp and a one-time value, so forged, replayed or stale reports are rejected.
  • If a secret leaks, the worst outcome is fake numbers for that one server. It gives no access to the server or to your account, and you can replace it in one click.

What you should know before installing

  • The install command needs sudo once, to create the user and the service.
  • With Docker, the agent's user joins the docker group to read the container list. Membership of that group is powerful on any system. The agent only sends read requests, and you can install with --no-docker to leave Docker out entirely.
  • Linux with systemd and Python 3.8 or newer. Windows and macOS are not supported.
  • It reports signs of problems. It is not an antivirus, a firewall or an intrusion detection system.

Questions

Frequently asked questions

The honest answers, before you sign up.

1

Can ismysiteup run a command on my server?

No. The agent has no code path that executes anything it receives. The reply to a report carries only a reporting interval and, when you delete the server, a signed removal notice.

2

Does it need a firewall rule?

No inbound rule. It needs outgoing HTTPS to ismysiteup.net.

3

How do I remove it?

Remove the server in the dashboard and the agent uninstalls itself at its next check-in. You can also run the uninstall command shown in the dashboard.