Automate Sentry triage
Turn incoming Sentry issues into scoped investigation and fix work.
Use a workflow to turn recurring Sentry triage into reviewable agent work, instead of someone opening the dashboard and reading through issues by hand.
Workflow shape
Connect Sentry once
From Settings → Integrations, connect your Sentry account (OAuth, no key to paste). Every workflow you build after that can read issues, events, and stack traces through the Sentry tool.
Trigger the workflow
Set a schedule (hourly works well for triage) or leave it manual and click Run now whenever you want a pass right away.
Have the agent investigate
Ask it to pull the newest or highest-frequency open issues, pick the ones worth investigating, and identify likely ownership.
Reproduce and fix
Have it run the relevant tests or reproduce the error locally, then propose a fix.
Review before merging
Open the run's workspace and review the diff and explanation like you would any other session.
Workflows run on a schedule or from manual dispatch today- there's no "trigger the moment Sentry fires" hook yet. Event-based triggers are on the roadmap, not in the product, so build this as a frequent scheduled pass (hourly, for example) instead of a per-issue webhook.
Sentry triage workflow editor
A workflow configured with an hourly schedule, a prompt describing the triage task, and the Sentry plugin enabled for tool access.
Keep the scope tight
Good triage workflows produce one of three outcomes:
- A small fix with evidence (the failing test, the fixed test, the stack trace it resolves).
- A reproduction and an owner recommendation.
- A clear "needs human investigation" handoff, with what was ruled out.
Tell the agent explicitly which of these is acceptable. A triage prompt that demands a merged fix every run will either produce weak fixes or silently do nothing on issues it can't resolve.
Reviewing runs
Every run shows up in the workflow's run history with its own workspace: the tool calls it made against Sentry, any file edits, and the test output all land in that session's transcript. Open a run the same way you'd open any other workspace and review it before merging.
Use team-wide workflows when triage should be shared across the whole engineering team instead of living under one person's account.