sentinelx-cloud-installer

The one-line installer for SentinelX. This is what https://get.sentinelx.app serves.

curl -fsSL https://get.sentinelx.app | sudo bash

Run that on any Linux host you want to operate from your LLM. The script clones sentinelx-cloud-core, sets it up as a systemd service, and walks you through enrollment.

Who this is for

Anyone who wants to install the SentinelX agent on a Linux server and connect it to Claude.ai or ChatGPT via the SentinelX hub.

What it does, step by step

1. Prereq check — confirms you're on Linux, running as root, with git, python3, and systemd available.
2. System user — creates an unprivileged sentinelx user that the agent will run as.
3. Sudo setup — by default, installs /etc/sudoers.d/sentinelx so the agent can run privileged commands without a password. The real security boundary is the allowlist in /etc/sentinelx/config.yaml — without that sudo rule, the LLM can't restart services or edit /etc/. To skip this step, set SENTINELX_SKIP_SUDO=1.
4. Clone & venv — fetches sentinelx-cloud-core to /opt/sentinelx-cloud-core and builds a virtualenv there.
5. Config skeleton — drops a starter /etc/sentinelx/config.yaml that you can edit later to control which commands the agent will allow.
6. Enrollment — opens an interactive enrollment flow:
   • Prints a URL like https://mcp.sentinelx.app/auth/dashboard/enroll?host_id=...
   • You open it in your browser, sign in with Google, copy the displayed enrollment token, paste it back into the installer.
7. systemd unit — installs and starts sentinelx-cloud-core.service. The agent connects out to mcp.sentinelx.app and stays connected.

After this completes, the host appears in your account on the SentinelX hub and you can target it from Claude.ai or ChatGPT.

Connecting to your LLM

In Claude.ai or ChatGPT:

1. Settings → Connectors → Add custom MCP
2. URL: https://mcp.sentinelx.app/mcp/mcp
3. Authorize with the same Google account you used during enrollment.

You're done. Try asking your LLM:

"List my SentinelX servers." "Show uptime and disk usage on my-vps."

Multiple servers

Run the installer on each one. Hosts are bound to your account, and the LLM can target them by:

• host_id — the unique ID generated at install time
• hostname — whatever the server reports (e.g. web-prod-01)
• label — a custom alias you set later via sentinel_set_host_label

If a name resolves to multiple hosts, the LLM is told which it is and asks you to disambiguate.

What you control

The agent only runs what's allowed in /etc/sentinelx/config.yaml:

• exec.allow — exact shell commands the LLM can run
• services.allow — systemd units the LLM can manage
• paths.allow_edit — files the LLM can read/write via sentinel_edit

A starter config is written at install time. Edit it to expand or restrict what your LLM can touch.

Manual install

If you'd rather not pipe a remote script into sudo:

curl -fsSL https://get.sentinelx.app/install.sh -o install.sh
less install.sh                     # read it
sudo bash install.sh                # then run it

The script is tiny (~7 KB) and stdlib-only.

Uninstall

sudo systemctl disable --now sentinelx-cloud-core
sudo rm -rf /opt/sentinelx-cloud-core
sudo rm -rf /etc/sentinelx
sudo rm -f /etc/sudoers.d/sentinelx
sudo userdel sentinelx
sudo rm /etc/systemd/system/sentinelx-cloud-core.service
sudo systemctl daemon-reload

The host stops appearing in your hub account within seconds (the WebSocket disconnects). Operational logs about that host roll off after 30 days.

Related

• sentinelx-cloud-core — the agent itself
• sentinelx-cloud-protocol — wire format spec
• Privacy Policy · Terms

License

Apache 2.0