Reflex Docs

Appearance

Getting started ​

Welcome to Reflex. This guide walks you through installing the agent, connecting your first server, and running your first deployment.

Prerequisites ​

  • A Linux server running Ubuntu 22.04+ or Debian 12+
  • Ability to use sudo only for steps that need elevated privileges (package install, systemd unit install) — prefer a non-root login and escalate only when the installer asks
  • A Reflex account — sign up here

Quick start ​

1. Install the agent ​

SSH into your server. Prefer download → inspect → run so you can review the script before it executes:

bash
curl -fsSL https://reflex.expertweb.tools/install/get-reflex.sh -o get-reflex.sh
less get-reflex.sh
# Optionally verify the agent binary checksum when the installer runs — see [Installing the agent](/getting-started/install#checksum-verification)
bash get-reflex.sh
Shortcut (less safe): pipe straight into bash

Piping a remote script into a shell skips review. Use only if you accept that risk:

bash
curl -fsSL https://reflex.expertweb.tools/install/get-reflex.sh | bash

The installer downloads the Reflex agent, registers it with your account, and opens a secure channel back to the Reflex control plane. For platforms, checksum pinning, and firewall details, see Installing the agent.

2. Verify the connection ​

Once installed, the agent reports back within seconds. Check your Reflex dashboard — your server should appear with a green "Connected" status.

3. Run your first deployment ​

Navigate to Pipelines in the dashboard, connect your Git repository, and trigger a deployment. Reflex handles:

  • Pulling your code from the configured branch
  • Running your build and release commands
  • Performing health checks after deployment
  • Rolling back automatically if health checks fail

4. Enable autonomous repair (what actually runs today) ​

Visit Playbooks in your server's settings. Reflex auto-executes these loops, gated by Brain confidence and team policy:

  • Pipeline health gates / rollback — failed post-deploy checks roll the release back
  • Node / Next.js keep-alive — restarts a named PM2 app or systemd unit after consecutive HTTP failures
  • cPanel service restart / disk cleanup — root-mode WHM hosts with a linked agent
  • Generic disk cleanup — stale tmp files, rotated logs, and old release dirs on any agent host
  • Generic systemd restart — allowlisted units only (nginx, PHP-FPM, MySQL/MariaDB, Redis, supervisor)
  • Generic OOM recovery — restarts allowlisted PHP-FPM when host memory is exhausted
  • Runaway process — SIGTERM for allowlisted processes after two high CPU/RSS samples

Other catalogue playbooks can be evaluated and suggested. See Repair playbooks.

Next steps ​

© Reflex by ExpertWeb. All rights reserved.