Get started

Start Tares

uv tool install tares     # or: brew install glassflow/tap/tares
tares up                  # choose 2) AI SRE

Add a model provider

The agents that watch and investigate need an AI model. Choose Add a provider on the home page and add your provider’s key. On Tares Cloud a model may already come with the workspace, and the row is ticked. Model providers

Connect your logs or metrics

Choose Connect and pick what you have: metrics (Prometheus), logs (Loki), alerts (Alertmanager or Prometheus alerts), or traces and logs (OpenTelemetry). Connect your data

No data at hand? Choose Try the demo service. It adds a demo service that you can break on purpose. On your local machine the demo service runs in Docker, and the page shows the two commands that start it.

See your services appear

Open Services. Each service your data talks about shows as one row with its state. If a service is missing or named oddly, see Services.

Turn on watching

Set up the watcher and the investigator with Guided setup. Open /projects/new on your Tares address (http://127.0.0.1:8787/projects/new  on your local machine), and say the goal in a sentence:

Watch my services every 10 minutes, and when something looks wrong, find the likely cause and write an incident note.

Pick Tares agents, choose Plan it, read the plan, and choose Looks right, set it up. Guided setup, Watching

Choose the Slack channel for notes

Connect Slack, then pick the channel under Settings, Notes. Incident notes land there, with buttons to take and resolve them.

Cause a test incident on the demo

With the demo service running, choose Open the demo and cause an incident. The investigator writes its note in about a minute, and it shows on Home, on Incidents and in Slack.