Sentry Error Triage Agent with OpenCode
The severity call is the whole product here, and how good it needs to be depends on how blunt your rules are. OpenCode lets you choose the model, so a team with sharp numeric thresholds can run something cheap while a team relying on judgement about known noisy errors runs something stronger. The rules, the codebase context and the Linear team survive the swap unchanged.
Watch it work before it's live
Run a staged conversation — no account needed. The agent handles it for real while a simulated world answers its tool calls; nothing touches real accounts, and nothing is actually sent.
TypeError: Cannot read properties of undefined (reading 'total')
Triggeredat renderOrderSummary (app/checkout/OrderSummary.tsx:118), at renderWithHooks (react-dom.production.min.js), at commitRootImpl. 4,120 events from 1,830 users in 40 minutes, all on release 2026.8.3 which shipped 52 minutes ago, every browser, cart and checkout pages only.
Set up in minutes
Using this template drops you into a guided setup. It asks exactly this, nothing else:
Connect Sentry
One sign-in. The agent acts through your account, scoped to what this template uses.
Connect Linear
One sign-in. The agent acts through your account, scoped to what this template uses.
Connect Slack
One sign-in. The agent acts through your account, scoped to what this template uses.
Severity rules
How you decide an error matters - the event and user counts, the time window, and how recent a release has to be for each level, plus what is severe enough to interrupt the channel.
Codebase context
What the services are and who owns which area - service names, the parts of the code they map to, and any known noisy errors you want scored down.
Runs on OpenCode
Preselected for this page — connect your OpenCode account during setup, or switch to NoClick's built-in models with one click.
Watch it handle a test run
A staged conversation against a simulated world — then it’s live.
Why OpenCode for this agent
Thresholds or judgement
Sharp rules leave little to interpret and a small model covers them. Vaguer ones lean on the model, and you get to decide which situation you are actually in.
Re-test on the same errors
The three staged errors are a fixed comparison, so changing model is something you can measure rather than something you guess at.
Before you fork
How do I know a cheaper model still scores correctly?
Run the staged post deploy spike, the quiet long tail and the single customer error, then read the verdict and the named rule in each ticket. The two quiet ones matter most, since over scoring them is how an alert channel stops being believed. If both land at the lower level with a stated reason, the cheaper model is doing the job.
Run it with a different agent
Put Sentry Error Triage Agent to work on OpenCode
Free to start. Guided setup, a test run against staged conversations, and it's live.