1. Connect your sources
1
Past incidents
Let investigations find similar incidents from your history and the fixes that worked before. See Past
incidents.
2
Channels
If you use Slack, add the channels where your team shares deploys, config changes, and incident context. See Slack
channels. Connecting channels as a source is available for Slack only.
Investigations in Microsoft Teams still read their own incident channel, but can’t draw on other channels.
3
Documentation
Sync your runbooks and reference docs from Confluence, Notion, GitHub, or GitLab. See
Documentation.
4
Code repositories
Connect GitHub or GitLab so investigations can link relevant pull requests and read your code. See Code
setup.
5
Telemetry
Connect the observability tools your team uses during incidents. See
Telemetry.
2. Choose when investigations run
Decide whether investigations run for every incident, only when conditions are met, when a workflow triggers them, or only on demand. See Triggering investigations for the options.3. Run your first investigation
Create a test incident and trigger an investigation to see it in action. In Slack, use/inc investigate in the incident channel. Otherwise, set investigations to run automatically or trigger one from a workflow.
What’s next
- See what an investigation can read: the conversation, the call, shared files, and your incident’s own details. See What we can see.
- Understand how we measure it: how we grade investigations for accuracy and use it to keep improving them. See Measuring accuracy.
- Try the chatbot: tag
@incidentin an incident channel to ask about logs, code changes, past incidents, or recent deploys. See the chatbot docs. - Share feedback: use the thumbs up/down buttons on investigation messages, and tell us what’s working and what isn’t.