Let Seer do more of your work in Sentry

[Dhrumil-Sentry]

TL;DR

• You already do a lot of manual work in Sentry (querying, investigation, triage)
• Seer already helps with issue RCA and fixes
• We’re experimenting with a new way to let Seer take on more of that work directly
• Looking for early volunteers willing to actively use it and give feedback - Just comment below or email seer@sentry.io with your org slug.
• If you have access, press Cmd + / anywhere in Sentry to try it

The work you already do in Sentry

If you use Sentry regularly, a lot of your time goes into manual work:

• Running queries across issues, traces, and logs
• Reading issues and trace waterfalls to understand what’s happening
• Triaging many issues and identifying problematic releases
• Investigating error or latency spikes after deploys

This work is necessary, but time-consuming.

What Seer already does today

Seer already helps by automatically root causing issues and proposing fixes when possible.

This happens by default for many issues, without manual intervention.

What’s changing

We’re experimenting with a new way to interact with Seer that lets you direct it to take on more of this work.

Instead of manually driving every query or investigation, you can ask Seer to handle more of the analysis and triage inside Sentry.

The goal is to reduce time spent navigating data and stitching together answers.

The kinds of work Seer can do for you

With this new interaction UX, Seer can take on work you’d otherwise do manually across Sentry. You can ask follow up questions and steer it - It has access to several tools and knows how to work through your data in Sentry.

Querying and summarization

Seer can translate natural language questions into concrete queries, run them across your data, and summarize the results.

This saves you from manually constructing queries, iterating on filters, and interpreting charts.

Investigations

When you see a spike on a chart or receive an alert, you can ask Seer to investigate directly.

Seer will:

• identify the relevant time window
• inspect related issues, spans, and traces
• narrow down likely causes and outliers

This replaces a series of manual queries and correlation steps.

Triage and prioritization

You can ask Seer to help with triage decisions, such as identifying who has the most context to fix an issue.

Seer can use stack traces, error context, recent code changes to guide prioritization.

This is early, and not everything will work well yet. We want to understand where this clearly saves time versus where it doesn’t.

UPDATE: Seer Agent is now in Open beta (4/26/26)

Seer Agent is in open beta for all Sentry users. Ask anything about what's happening in your application and get answers based on all the telemetry and context Sentry knows about it. Ask it about anything you'd normally spend hours digging through like why your p95 latency spiked like why your p95 latency spiked, what's causing that crash loop, or why a specific user keeps hitting errors. Check out the docs for more.

How to try it

Open any page in Sentry, hit Cmd + / or click the “Ask Seer” button, and ask it something. Or if you have our Slack integration, type a question @sentry

Drop feedback in the discussion or Email us at seer@sentry.io

We’re especially interested in:

• Where this clearly saves you time
• Where you don’t trust the results
• What work you wish Seer would take on next

[cquanu]

Emailed

[Dhrumil-Sentry]

@cquanu You should have access now - Just use Cmd+/ on to access the UX - We'd love to hear any feedback 🙏

[alaney2]

would love to test

[Dhrumil-Sentry]

@alaney2 We've enabled this for your org - Cmd+ / on any Sentry page to access this ! Feedback much appreciated 🙏

[alaney2]

How does this work? I'm pressing Cmd / on https://athenahq.sentry.io/issues/ and not seeing anything

[Dhrumil-Sentry]

@alaney2 you org has disabled the ability to show Gen AI features - If you go to your org settings and enable this then Cmd + / will work

[alaney2]

awesome thank you!

[cpritchett]

Would love to test. Org - blavity

[Dhrumil-Sentry]

@cpritchett You should have access ! Just try "Cmd + /"

[dirq]

+1 on joining the Alpha.

[Dhrumil-Sentry]

@dirq You should have access - We added a UI entrypoint to access this UX too

[dirq]

Thank you, I'll check it out soon.

[EllAchE]

Have been trying a lot of PR bots (did power rankings of the ones I tried here https://loganharless.com/blog/ai-pr-bot-rankings), happy to add another to the list 😂

[Dhrumil-Sentry]

@EllAchE Error prediction in code review does not need any special access flagging - You can already try it yourself - Here's the blog for this

[EllAchE]

thought i saw comments on PRs advertised somewhere

[aliu39]

Sharing some useful queries to try:

• summarize what happens in this trace

  • (for me trace waterfalls take a while to read so getting a quick summary is great)
  • explain the most relevant services/code the trace passes through
  • expand+summarize what happens under a specific span
  • summarize errors, log msgs, metrics in the trace... dig into specific issue details, replay summary, log/metric tags from there.
  • Whatever you can imagine doing in the UI, try "vibe-ifying" it by asking chat, then following the links

• what causes this issue? are there similar or related errors? summarize related replay? can you propose code changes for it (interactive autofix)

• summarize issues in my project that are new, escalating or regressed (optionally matching some criteria). Include their impact and description of when things happened

  • ask opinion on which are highest priority/easy fixes

• what causes this spike (or other trends) in errors/span count at this time range? - for an issue, project, txn...

  • this one's still pretty work-in-progress so feedback's very welcome

[niebloomj]

Would love to test! Org is https://github.com/frumfinance

[kingdomseed]

Yeah I'd love to be on the test team for this: jason-holt-digital-llc

[aniruddhaadak80]

Expanding Seer to automatically suggest code fixes alongside the trace is a huge step up. It would be amazing if Seer could eventually auto-generate unit tests that reproduce the specific production exceptions it finds, effectively proving the issue exists before presenting the fix.