Sentry
Sentry · Analytics Platform

AI for Sentry.

Ask Sentry what broke – not how to write the query.

Read-only · Flexible connection options

Read-only query
> Which unresolved production issues affected the most users in the last 7 days?
✓ Read-only · APISetup time varies by connection

Turn Sentry's issues, events, projects, and releases into direct answers.

Ask which errors regressed after a deploy, what's burning the most users, or where the team keeps paying the same reliability tax. AI for Database queries the Sentry API in read-only mode and returns the evidence – without making you wrestle with filters, exports, or another dashboard.

Sentry stays the system of record. AI for Database gives you a faster way to interrogate it.

What data you can query

Issues

Grouped errors with titles, severity, occurrence counts, and user impact

idtitlestatuslevelcountuser_countfirst_seenlast_seenpriority

Events

Individual error occurrences with stack traces, breadcrumbs, and user context

event_idmessageculprittimestampplatformenvironmentrelease

Projects

All projects in your org with platform, status, and metadata

slugnamestatusplatformdate_createdfirst_event_date

Releases

Deployment history with version, commits, and associated projects

versiondate_createddate_releasedproject_countcommit_count

Problems this solves

The issue list is not a priority list

Sentry can show hundreds of active issues. It can't decide which ten deserve engineering time before the next planning meeting.

Release regressions hide inside familiar noise

A deploy lands, event volume climbs, and the new failure gets mixed in with errors everyone has already learned to ignore.

Simple questions turn into filter archaeology

Comparing projects, releases, environments, and time windows should take seconds. Too often it means reconstructing the right search syntax and clicking through multiple views.

Reliability context gets lost between incidents

The team fixes the obvious fire, but recurring exceptions, noisy services, and long-running regressions quietly become part of the production baseline.

How AI for Database helps

Triage by impact, not inbox order

Rank unresolved issues by event volume, affected users, recency, project, or release so the team starts with failures that matter.

> Which unresolved production issues affected the most users in the last 7 days?

Interrogate every release

Compare error activity before and after a release and surface issues that first appeared after deployment.

> What new or reappearing issues showed up after the latest release of the API project?

Cross-project questions in plain English

Query multiple Sentry projects without rebuilding the same filters or exporting data into a spreadsheet.

> Which three projects produced the most error events this month, and what were the top issues in each?

Turn incident review into evidence

Pull the issue and event history needed for postmortems, weekly reliability reviews, and leadership updates.

> Show me the top 5 issues by user count that are still open, with their first and last seen timestamps

Expose recurring engineering debt

Find old unresolved issues, repeated regressions, and projects that generate a disproportionate share of production errors.

> Which issues have increased by more than 50% in event volume since the previous release?

Example queries

> Which unresolved production issues affected the most users in the last 7 days?

Filters open issues by environment, sorts by user_count descending, compares with previous 7-day window

> What new issues showed up after the latest release of the API project?

Compares issue first_seen timestamps against the latest release date for the API project

> Which issues have increased by more than 50% in event volume since the previous release?

Aggregates event counts per issue before and after the previous release, calculates percentage change

> Show me unresolved issues older than 30 days that still generated events this week

Filters issues by status=open and first_seen > 30 days ago, checks for recent event activity, ranks by affected users

> Which three projects produced the most error events this month?

Aggregates event counts per project for the current month, orders descending, returns top 3 with contributing issues

Dashboard templates

Release Regression Watch

Track new and reappearing issues around every deploy

  • New issues by release (bar)
  • Event-volume change from previous release (table)
  • Affected users per release (line)
  • Projects with sharpest post-deploy increase (highlight)

Reliability Debt Board

Surface what's been ignored too long

  • Top unresolved issues by age and impact (table)
  • Recurring issues across releases (list)
  • Error volume by project (bar)
  • High-frequency failures still unassigned (KPI)

Automated workflows

Morning production triage

Unresolved issues that grew fastest in the last 24 hours, ranked by user impact

Weekdays at 8:30 AM

Post-deploy check

Newly introduced or reappearing issues after the latest release, with event and user impact

After every deploy

Weekly engineering review

Which projects produced the most errors, which issues repeated, and what remained unresolved

Weekly on Friday at 4 PM

Incident follow-up

When an issue started, how volume changed, which releases overlapped, and whether similar failures appeared elsewhere

On-demand after incidents

Key metrics

Unresolved issue countError events over timeUsers affected by issueNew issues by releaseRegressed or reappearing issuesEvent-volume change after deploymentMean and median issue ageError volume by projectRepeat issues across releasesHighest-impact unresolved issues

Connection options

Direct API

Connect via Sentry's REST API with a read-only auth token. No warehouse, no ETL, no infrastructure – just your API token and org slug.

Pros

  • + Setup in under 2 minutes
  • + Real-time data
  • + No additional infrastructure
  • + Read-only – zero risk to your Sentry account

Cons

  • - Rate limited (~1000 requests/min – more than enough for queries)
  • - No historical backfill beyond Sentry's own retention

Setup

01
Create an API tokenIn Sentry, go to Settings → Auth Tokens → Create New Token. Select 'Project: Read' and 'Org: Read' scopes
02
Grab your org slugOpen your Sentry dashboard. The slug is in the URL: sentry.io/{org-slug}/
03
Connect in AI for DatabaseAdd a new connection, select Sentry, paste your token and org slug

How to get your Sentry API token

01

Log in to sentry.io and navigate to Settings → Auth Tokens

02

Click 'Create New Token' and name it (e.g., 'AI for Database')

03

Under scopes, check 'Project: Read' and 'Org: Read' – that's all you need

04

Click 'Create Token' and copy it immediately (it starts with 'snr_')

05

Grab your org slug from your Sentry URL (the part after sentry.io/)

Tip: This token is read-only. It can't create, modify, or delete anything in your Sentry account.

View official Sentry documentation →

FAQ

Does this replace Sentry?

No. Sentry stays the system of record for error and event data. AI for Database gives you a faster, conversational way to query that data and combine results across issues, projects, releases, and time windows.

Can it change issue status, assign owners, or modify alerts?

No. The integration is read-only. It queries data through the Sentry REST API and does not resolve issues, change project settings, or edit alert rules.

What Sentry data can I ask about?

Issues, events, projects, and releases. Available detail depends on the API permissions and data in your Sentry account.

Will it tell me the root cause of an error?

It can surface patterns and evidence – when an issue appeared, which release preceded it, how event volume changed. It doesn't inspect your code or guarantee a root-cause diagnosis.

Your Sentry speaks English now.

Free plan available · No credit card required