Quickstart

Install 3AM on one Linux server with a free trial, open the console, connect Prometheus and Slack, and see your first incident. About 30 minutes.

By the end of this page, 3AM is running in shadow mode: it reads your alerts, proves causes and writes down the fix it would make, but changes nothing. That's the safe way to see what it does on your systems.

You need:

  • A Linux server you can SSH into. 8 vCPU and 32 GB is what we run; a 2 vCPU, 4 GB VM is enough to try it.
  • An OpenRouter API key, if the server has no GPU. You can use your own model endpoint instead.
  • A Prometheus 3AM can reach, and a Slack workspace where you can add a bot.

Run the installer with a trial

On the server, as your normal user (not root):

curl -fsSL https://get.3am.si/install | sh -s -- --trial

It asks a few questions, in this order:

  1. Where should 3AM run? Choose 1, this machine.
  2. Install Podman? Only if neither Podman nor Docker is installed. Say yes.
  3. Your trial. Your name, work email, company, role and a little about your estate. A trial code arrives by email within a minute: paste it in. Already have a code? Use --code 3AM-TRIAL-… instead of --trial.
  4. Where the models run. Only asked when the server has no GPU. Choose 2 and paste your OpenRouter key (it isn't shown as you type), or 1 for your own endpoint.
  5. How you'll open the console. Choose 1, an SSH tunnel. Nothing on the server is exposed.
  6. Install now? Type y.

It then downloads 3AM, checks every file's signature, and starts it. It ends like this:

Output
==> 3AM is running
  Console      http://localhost:8700, from your laptop through an SSH tunnel:
               ssh -N -L 8700:127.0.0.1:8700 ana@203.0.113.10
  Admin token  Q3xk9…   (shown once; also in the 3am-data volume at admin-token)
  Next         sign in to the console with the admin token: the trial starts by itself, then connect your tools

Copy the admin token. You need it once, to create the first owner.

Open the console

On your laptop, run the ssh command the installer printed, and leave it running:

ssh -N -L 8700:127.0.0.1:8700 ana@203.0.113.10

Open http://localhost:8700. Enter the admin token and your email: that makes you the first owner, and starts the trial.

Connect Prometheus and Slack

The setup wizard opens. Under Monitoring, choose Prometheus / Alertmanager and enter its address. Test connection runs a real query before anything is saved:

Test connection
Connected. Prometheus 3.15.0 + Alertmanager 0.34.1

Under Approvals, add Slack: a bot token, and the channel where 3AM should ask. The Prometheus and Slack guides list exactly what access to give.

Then press Discover. 3AM reads your scrape jobs and alert rules, works out your services, and turns on the checks that fit them. Leave autonomy at Shadow.

Send a test alert

Add a rule to Prometheus that always fires, and reload Prometheus:

groups:
  - name: 3am-test
    rules:
      - alert: ThreeAMPipelineTest
        expr: vector(1)
        labels: { severity: critical, service: monitoring }

Within a minute it appears under Incidents. A test alert has no real cause, so 3AM says so and hands it to a person. That's the right answer. Delete the rule when you're done.

If something goes wrong

Next

  • Leave it in shadow for a week. The Digest compares what 3AM would have done with what your team did, and says when a service is ready for approvals.
  • Connect what 3AM can fix: Kubernetes, databases and servers.
  • Bring your team in with single sign-on: Sign-in.

On this page