Skip to content
TriageAI

Bug triage

From bug report to root cause in 60 seconds

Give TriageAI a Jira ticket, Sentry error, GitHub issue or raw stack trace. It investigates your codebase like a senior engineer and returns the root cause, a fix and a regression test.

Bug report, from Sentry

TypeError: Cannot read properties of undefined (reading 'amount')
    at applyDiscount (src/billing/discount.ts:42:31)
    at calculateTotal (src/billing/invoice.ts:88:18)
    at POST /api/checkout (src/routes/checkout.ts:27:22)

Root cause

src/billing/discount.ts:42 reads coupon.rules[0].amount, but commit a41c9e2 made coupon rules optional. Coupons without rules crash checkout.

Suggested fix

- const off = coupon.rules[0].amount;+ const off = coupon.rules?.[0]?.amount ?? 0;

Regression test

it("ignores coupons with no rules", () => {
  const coupon = { code: "WELCOME", rules: [] };
  expect(applyDiscount(100, coupon)).toBe(100);
});
Example diagnosis. Review before merging.

Bug triage eats the time you meant to spend fixing

  • 45 to 90 minutes per bug, before anyone writes a fix

    Reading the trace, finding the code, working out what changed and who changed it. The same investigation every time, done by hand.

  • Junior engineers get blocked

    Without the context of an unfamiliar codebase, they wait for a senior engineer, who then stops what they were doing.

  • Fixes ship without regression tests

    Under pressure the test is the first thing to go, so the same bug comes back a few releases later.

Automated defect triage that ends in three things you can use

Every diagnosis comes back in the same structure, posted to the ticket it came from.

Fewer hours lost to triage

Take root cause from an hour to a minute and a 10-person team reclaims 20+ hours per sprint.*

Estimated impact of bug triage before and after TriageAI
MetricBeforeWith TriageAI
Time to root cause45–90 min~60 sec
Fixes shipped with a regression test~30%100%
Junior engineer self-sufficiencyLowHigh
Senior engineer interruptionsFrequentMinimal

* All figures on this page are estimates, not measured customer results. Actual results depend on codebase size, bug type and team.

The last step of every outage analysis

When an outage traces back to code, outage analysis hands off to bug triage, so the incident ends with a fix and a test instead of a rollback and a to-do.

See outage analysis
How does it avoid making things up?

The agent ties every claim to a file, line or commit it actually read, and says so when it can't reach a confident answer. The investigation trace shows every tool call it made, so you can check its work.

Does it push fixes or merge code?

No. TriageAI posts a diagnosis as a comment. Every result is a hypothesis for an engineer to review, and there is no auto-merge anywhere.

Which languages are supported?

The demo investigates a JavaScript codebase. Multi-language support for Java, Python, Go and TypeScript is on the roadmap.

Which model does it use?

Claude, from Anthropic. The model is configurable per deployment.

Triage your next bug in about a minute

Paste a stack trace or a bug report and get a root cause, a fix and a regression test back. Review it, then ship it.