1. Create your organisation
Register an account, then create an organisation from the dashboard. Every organisation is fully isolated — its own members, connectors, incidents, and audit log. Nothing crosses between them.
2. Connect your tools
Go to Connectors and add one of each: Grafana (where the alert and its context come from), GitHub (the repository the failing service is built from), and Jira (where the packaged incident lands). Each has its own setup guide in this section.
Use Test connection on each one before routing any alerts. A connector that has never been tested will simply contribute nothing when an incident arrives, and you would rather find that out now than at 3am.
3. Route an alert
Each connector shows a webhook URL. Point your alerting system at it and set the shared-secret header your administrator configured. Any JSON body works — the payload is stored verbatim, and the pipeline reads the original rather than a normalised copy.
- Copy the webhook URL from your Grafana connector
- Add it as a contact point (Grafana) or receiver (Alertmanager)
- Add the X-Incident-Copilot-Secret header with your shared secret
- Fire a test alert and watch it appear under Incidents
4. Review the first incident
Open it from the Incidents list. You get the root-cause hypothesis, how confident the model is in it, the evidence it was derived from, and the proposed fix. If the fix is a code change, there is a draft pull request waiting in your repository. If it is an operational change, there is a list of commands — which have not been run.