I'll be joining @sergical at @sentry on October 22nd to talk about connecting error monitors to coding agents! It's how we've closed the loop in our own software factory to have agents assign themselves work automatically. The flow we're using right now: - Connect to an alert webhook like Sentry - When errors happen, start a cloud agent run with the alert and the repo as context - Have the agent ping humans for code review (or better: with an HTML summary of the problem and the solution) - Merge and keep monitoring This closes the loop on reducing software bugs. We'll show you how to deploy it in-code yourself!
brandCooling
Sentry on X
Sentry is talked about warmly: plain praise and enthusiasm for its agent and MCP tooling are level at the top of the reception, while the biggest complaint is that self-hosting is heavy and can cost more than the SaaS.
First written 27 Oct · what changed · from 65 posts by 57 people
Tracking posts that say “@sentry”
Track this for your marketPosts, last 24h
0
−100%vs. previous 24h
Posts, last 7 days
198
45 in the last 72h
Velocity
28/day
7-day average
Momentum
−100%
Last 24h vs. a normal day this week
Posts per day, last 30 days
as of Nov 3, 09:00 UTCWhat changed
as of 27 Oct · checked once a monthFirst written 27 Oct. When what people say changes, it shows here.
Tone· 38 posts from 33 people· written 27 Oct
How do people feel about Sentry?
Reception is warm: plain praise (15 people) and enthusiasm for the agent and MCP tooling (14 people) are level at the top of the 33 people talking about it, and few give a negative verdict (3 people).
Plain thanks are one half of the positive talk: people credit Sentry with catching critical bugs so fixes come faster, with backing open source maintainers in Africa, and one calls it a "godsend" for WebGL failure modes on low-end devices that cannot be tested for ahead of time. The other half, level with it, is the agent-facing tooling: people describe wiring Sentry alerts into coding agents so the agent picks up the error and opens a PR, use the MCP server and its Claude Code plugin, recommend the new CLI as "perfect for my agent", and use Seer on trace data to get a quantified breakdown of where users hit capability gaps.
Smaller voices: 3 people say it works well when sized lean — a smaller machine, errors only, self-hosted with analytics and logs stripped out — and 3 give a negative overall verdict, calling it clunky or claiming it is one prompt away from being irrelevant.
- Loves the AI and MCP tooling17
- Praises it15
- Runs fine if you tune it3
- Disappointed or dismissive3
See 5 of 38 postsHide the posts
While everyone is showing off their factories/loops/.... you can also just use Claude and your @sentry trace/error/telemetry data in MCP and make things faster Be like @scttcper
Next up in the gratitude series: Sentry. Every critical bug they've helped us capture has made Sequence more reliable, and we've fixed issues faster because of it. Thank you, @sentry!
holy hell. @sentry is a godsend for developing cross-platform browser games. WebGL has *all kinds* of failure modes on low-end devices that LLMs are not good at predicting ad hoc & are impossible to test for ahead of time would be flying blind w/o it. my opinion now is actually that WebGL games were ~impossible to mass market before LLMs bc device surface area = too expensive to get 100% coverage; too many bad experiences
You can just use a smaller machine it still works + if you only want errors then its also way smaller to host. It replaced all of my observability so its worth it unless you dont care about this for which errors-only is for link
5 of the 38 posts behind this answer, one per person. The other 33, and every new one as it’s posted, come with tracking. Get early access
Pain points· 19 posts from 18 people· written 27 Oct
What do people complain about with Sentry?
Self-hosting is the biggest complaint, from 7 people: it eats huge resources and can cost more than the SaaS. Day-to-day roughness is next (4), then missing newsletter emails (3), nagging alerts (2) and trust doubts (2).
On self-hosting, people report a 32GB VPS dying twice in a day while running only crash reporting, and a Helm HA install that "burned more resources than the production workloads it was monitoring"; one moved to the SaaS because the infrastructure cost more than the subscription, and others suspect the bloat is deliberate to push people onto the hosted product.
The clunky complaints are about daily feel: a UI described as "soooo clunky", glitches, sign-in mixed with a second factor, and Sentry.js multiplying app.js size several times over.
Three people say they signed up repeatedly for the newsletter and it never arrives — not even in spam. Two say alerts nag them about breakage they already know about or have not had time to fix. On trust, one warns that JavaScript agent tracing stores model inputs and outputs by default through dataCollection.genAI, and one would not recommend Sentry over the conduct of a staff member.
- Self-hosting is heavy and expensive8
- Clunky and glitchy4
- Newsletter emails never arrive3
- Alerts feel like nagging2
- Trust worries2
See 5 of 19 postsHide the posts
This instance is only for Sentry, with nothing else on it, so this is totally Sentry's fault here.
Self-hosted Sentry via Helm in HA mode just wasn't sustainable for us. It ended up burning more resources than the production workloads it was monitoring. On top of that, small gaps in the Helm chart like the lack of native IRSA support for AWS S3 meant jumping through extra security hoops, forcing us to manage dedicated IAM service users and static access keys instead of using fine-grained pod-level IAM roles.
That’s their entire business model. Make it so difficult (and expensive) for self-hosted people that you have to just use their saas.
Last year I did a migration from self hosted Sentry to SaaS Sentry, because the infrastructure for hosting it was more expensive than their SaaS subscription. Makes no sense.
Yes, please do, sentry is soooo clunky
5 of the 19 posts behind this answer, one per person. The other 14, and every new one as it’s posted, come with tracking. Get early access
Alternatives· 6 posts· written 27 Oct
What are the alternatives to Sentry?
This is thin ground: the largest group is Cloudflare, weighed by 2 people, and each of the other alternatives — SigNoz, GlitchTip, Rustrak and the LGMT stack — is named by one person.
The Cloudflare pair split: one is glad they can finally "remove sentry" now Cloudflare covers it, the other prefers Sentry's CLI for their agent and says getting basic HTTP analytics out of Cloudflare means falling back on its REST API.
Several of the other mentions come out of self-hosting frustration. One person runs GlitchTip, the free self-hosted option, and still wants an up-to-date lightweight open source version; another gave up on self-hosted Sentry for the LGMT stack of logs, metrics, traces and alerting, allowing it is not as fancy for errors. One weighs Rustrak and lists what they would lose — grouping by message rather than stack trace, alert rules whose conditions are saved but never checked, weaker tag search. One simply suggests SigNoz.
- Cloudflare2
- SigNoz1
- GlitchTip1
- Rustrak1
- LGMT stack1
See 5 of 6 postsHide the posts
I think @sentry CLI is perfect for this. They have MCP and all the other AI jazz, but their current CLI is already perfect for my agent. With @Cloudflare, even getting basic HTTP analytics for my website isn’t possible through their MCP or other AI tools. I have to use their REST API instead. Not the best experience.
Omg finally can remove sentry @sentry
I've been looking into this too. It's crazy how much resources it uses. I've tried Glitchtip which is free selfhostd which works well. But I really want an up to date open source version that is lightweight.
I gave up on self hosted Sentry and just went for the LGMT stack. Okay it’s not quite as fancy as Sentry for errors but you get logs, metrics, traces and alerting. Plus with an agent and the gcx cli, you can diagnose anything with all that data.
Here’s what my agent thinks, did it get it right @roberteggl? What we'd lose compared with Sentry: •Grouping by message, not stack trace. Rustrak groups on exception type plus message. It strips numbers, UUIDs and hashes, but not file paths, so a panic whose message names a plugin path would open one issue per plugin. We'd need client-side fingerprints to fix that. •Alert filters. Alert rules have a conditions field that is saved but never checked. Alerts fire on new issue, regression or unmute, with no filter on environment, level or tag. The !test_event:true exclusion in docs/error-reporting.md can't be set up; muting the test-report issue is the workaround. •Tag search. Issue search is free text over the title, culprit and transaction, so you can't search for something like arch:x86_64.
5 of the 6 posts behind this answer, one per person. The other 1, and every new one as it’s posted, come with tracking. Get early access
Questions· 4 posts· written 27 Oct
What do people ask about Sentry?
Questions are sparse and practical: 2 people ask what the catch is or what the biggest hassle with self-hosting was, alongside one asking whether symbol upload is automatic and one asking how real traces become eval cases.
A happy user wonders whether anyone has had a bad experience, and a reply asks another user what the biggest hassle with self-hosting was. The two questions asked are both about the downside, and one is specifically about self-hosted setups.
The other two are about the agent workflow. One asks whether an agent that sets up crash reporting from the editor handles symbol upload automatically on each build or whether that step is still manual. One, looking at a session on traces, wants to know how you decide which failures become regression cases rather than production alerts.
- What's the catch?2
- Is symbol upload automatic?1
- How do traces become eval cases?1
See the 4 postsHide the posts
Have been a happy @sentry user, wondering if anyone has had a bad experience
what was the biggest hassle with self hosting?
having the agent set up crash reporting directly from the editor is smart. does it handle symbol upload automatically on each build, or is that still manual?
This sounds worth joining. Turning real traces into eval scenarios is exactly the bridge I’m trying to understand better, especially how you decide which failures become regression cases versus production alerts.
Wishes· 2 posts· written 27 Oct
What do people want from Sentry?
Two people, two wishes: put the app version in the alert, and share a demo repo at a fixed commit so a live-session workflow can be repeated.
One wants an alert to carry the app version that produced it, so a fixer — human or agent — is not sent chasing bugs in code that has already changed.
The other asks for the demo repo from a Sentry-to-Warp session at a fixed commit, plus the steps to reproduce the issue, so attendees can repeat the workflow afterwards and pinpoint where they get stuck.
- Put the app version in alerts1
- Share a reproducible demo repo1
See the 2 postsHide the posts
Including the app version that produced the alert would save the agent from chasing bugs in code that has already changed.
A useful take-home would be the demo repo at a fixed commit plus the steps to reproduce the issue. That would let attendees repeat the Sentry-to-Warp workflow after the live session and pinpoint where they get stuck.
In their words
How do people talk about Sentry?
The phrases people use most, counted once per person.
- 1“self hosted”3
- 2“self hosted sentry”2
- 3“crash reporting”2
- 4“symbol upload”2
- 5“agent tracing”2
- 6“mcp server”2
- 7“sentry for errors”2
Posts are shown the way X shows them. One of them yours? Here’s why it’s shown, and how to take it off.
Also in Observability and error monitoring
- Datadog+14%The largest group on the page is praise, 12 of 53 people, many of them wiring Datadog alerts and traces into agent workflows; the two complaints raised as often as each other are cost and sensitive data ending up in Datadog.245 posts / 7d
- Grafana Labs+31%Grafana's biggest theme is 39 people running it in their own stack who say it does the job, ahead of 9 who love it loudly.213 posts / 7d
Get every post about Sentry, filtered and labelled
This page is the summary. Tracking gives you the stream behind it for any brand, keyword or @mention: each new pain point, request and question as it’s posted, with the spam and promotions already taken out.