Connect your data
The AI SRE sees what you connect. Open Settings, Tools to see what is connected and to add more.
| What | Connector | What it gives the agents |
|---|---|---|
| Metrics | Prometheus | Error rates, latency and traffic per service. |
| Alerts | Alertmanager, Prometheus alerts | What your alerting already fired. |
| Traces and logs | OpenTelemetry | Spans and logs your services send over OTLP. |
| Logs | Loki, Docker logs, Vercel | The log lines around each change. |
| Deploys | GitHub | Commits and releases, shown on the chart of the service they belong to. |
| Anything with an API | HTTP poll, Webhook | Status pages, job results, your own events. |
Your coding agent can set up a connector for you over MCP. Set up sources with an agent
Tie events to a service
Every event is counted for the service it names. Tares reads the service from the label called
service, or else from the connector’s main label, such as a Prometheus job. Keep the same
service name across your metrics, logs and alerts, so all of them land on one service.
If one service shows under two names, merge them on the service’s page. If a service shows that is not one, hide it. Services
Link repos to services
On a service’s page, link the GitHub repos that deploy it. Its deploys then show on the service’s chart and timeline, and the investigator sees them next to the errors that followed.