Key findings
- The other 8 code review accounts describe wrong findings or delays, including a review job that blocked a merge.
- Wrong findings cost teams time. 6 accounts raise them, including two findings Seer rated high and critical that a City of Helsinki developer rejected on one pull request.
- All 9 accounts that discuss platform and setup limits report problems. They cover Bitbucket and Azure DevOps, GitLab before July 2026, an 8-repository limit, and handoff settings that did not save.
- Seven accounts give an opinion on how far Seer should act on its own. Four let it run, and three stop it before it opens pull requests.
- Four accounts discuss price and two object to it. One team went over budget on legacy per-run pricing, and one project chose its own pipeline over the US$40 per contributor charge.
- Two companies wrote up Seer on their own blogs. Visualsoft merged several Seer fixes in a 2025 trial, and Dwayo AI called off its rollout in October 2026 after CVE-2026-90999.
Seer's code review earns the most praise from engineers, with real catches in 15 of the 23 public accounts that discuss it, most fixed before merge. The complaints are platform limits and wrong findings, and teams split on how far to let Seer act alone.
Who is writing about Seer
Most first-hand writing about Seer sits on GitHub. Of the 47 accounts, 40 are GitHub records, 17 of them issues and 23 pull requests or review replies. The rest are 3 X posts, 1 Hacker News comment, 1 piece of security research, 1 company blog post, and 1 podcast transcript.
Two companies have described Seer on their own blogs. In July 2025, Jon Purvis, AI lead at Visualsoft, called Seer "a fantastic product" and wrote that his team "merged multiple Seer-generated fixes" after a person checked each one. He asked Sentry for pull requests that follow the team's own templates and for fixes of high-actionability issues with no manual start. In October 2026, Devji, co-founder of Dwayo AI, wrote that the team "decided not to proceed with the rollout" after CERT/CC disclosed CVE-2026-90999.
Code review draws the most first-hand detail
Engineers' replies on pull requests show Seer's review catching real bugs and flagging code that was fine. AutoGPT maintainers fixed a high-severity finding about reading text files without a size check. They rejected another as a verified false positive after checking GitHub's documentation. An Aperant contributor called Seer's path bug the one critical fix among several AI reviewers. An Exactly engineer said the flagged code was intentional.
Engineers at MIT Open Learning, Topsort, and Twenty fixed Seer findings, including a sensor cursor that would have dropped failures permanently. City of Helsinki and Give Lively developers rejected high or critical findings. An Oak National Academy developer doubted a finding until a colleague confirmed it in the docs. Several engineers accepted a finding and wrote a different fix.
One project made Seer's check a required review gate alongside Codex after Copilot's monthly review credits kept running out. Another team found Seer's reviews slow and asked to run them in the CLI before a commit. On GitLab, a Seer review job hung with no way to cancel or retry it, and the developer was blocked from merging.
Where teams hit limits
Seer supports GitHub.com and GitLab.com only. Bitbucket users have requested support since May 2025. In June 2026, an Azure DevOps user said their team would not buy without it. Both requests were still open on October 25, 2026. A medical software team on GitHub Enterprise lacked AI Code Review in late 2025.
Sentry made GitLab support generally available in July 2026. Earlier, in May, a GitLab team canceled after Seer required a connected repository to analyze issues. Another team spent about a developer-day reading Sentry's source to configure Autofix during the GitLab beta.
In one pnpm repository, Autofix added a package-lock.json to every pull request. Handoff settings failed to save at the organization level. Other users report that Autofix commits cannot be signed and that Seer opened pull requests without a Seer subscription.
Price and running your own agent
Two accounts compare Seer with an agent they control. Shepherd's maintainers replaced Seer with their own fix pipeline, citing its closed source, its US$40 per active contributor price, and its ignorance of their house rules. A Hacker News commenter found Seer unnecessary because they can send Sentry data straight to Claude.
A Deliverect QA lead went over budget on legacy per-run pricing after setting automation too aggressively, and was locked out of Seer settings until the next billing period.
How far to let it run
Four accounts let Seer act alone or ask for more of that. One developer set Autofix to always run on one app, and another asked for autonomous fixes of high-priority issues. Visualsoft's team asked Seer to fix high-actionability issues with no manual start. Three stop earlier. The Zcash Foundation plans a root-cause-only pilot on a public repository, and another team objected to a pull request nobody requested. The PhantomFix research from agyn recommends pausing automatic handoff until teams have reviewed it or a fix is available.
Methodology
GILD reviewed 47 public Seer accounts, published April 2025 through October 2026, from October 12 to October 27, 2026: GitHub issues, pull requests and review replies, X posts, a Hacker News comment, security research, a company blog post, and a podcast transcript. Searches covered GitHub issues and pull requests, Hacker News through its Algolia API, and Reddit through the Arctic Shift archive. GILD capped GitHub issues at 17, about a third, because issues lean toward missing or broken behavior. It excluded Sentry employees, Sentry's blog and customer pages, and contest entries Sentry sponsored. Sentry's customer stories about Bolt and Honra carry no publish date and are left out. On October 25, 2026, GILD searched company engineering blogs again and found Visualsoft's and Dwayo AI's posts. Visualsoft's post counts among the 47 accounts, and this record quotes Dwayo AI's outside them. Every count describes the accounts GILD reviewed.
Products in this piece
Sources
- Public source
- 20
- Official documentation
- 5
- AutoGPT pull request #12616: a Seer finding resolved as a verified false positive (Jun 25, 2026)
- AutoGPT pull request #12125: a Seer finding fixed (Feb 17, 2026)
- CloudHub-Social Charm pull request #105: review gate requires Seer and Codex (Jul 9, 2026)
- getsentry/toolkit issue #900: Seer code review in the CLI or IDE (Apr 14, 2026)
- getsentry/sentry issue #92317: Seer Bitbucket integration (May 27, 2025)
- getsentry/sentry issue #93724: Seer GitLab integration (Jun 17, 2025 to Jul 23, 2026)
- getsentry/sentry issue #95796: Seer support for Azure DevOps (Jul 17, 2025)
- getsentry/sentry issue #117489: Autofix repository limit (Jun 12, 2026)
- getsentry/sentry issue #125174: Autofix lockfile generation (Sep 22, 2026)
- getsentry/sentry issue #121146: Sign commits with Seer Autofix (Aug 4, 2026)
- City of Helsinki linkedcomponents-ui pull request #646: two Seer findings rejected as not bugs (Sep 28, 2026)
- MIT Open Learning ol-data-platform pull request #2518: Seer findings fixed and one declined (Aug 7, 2026)
- getsentry/sentry issue #125858: Seer opening pull requests without a subscription (Sep 29, 2026)
- getsentry/sentry issue #107592: Cannot configure Seer when out of budget (Feb 4, 2026)
- Zcash Foundation Zebra issue #11635: pilot Seer root cause analysis without automatic pull requests (Oct 9, 2026)
- erwins-enkel Shepherd issue #2467: Sentry auto-fix as a Seer replacement (Sep 24, 2026)
- jblake on Hacker News: Seer is interesting but unnecessary next to Claude (Apr 19, 2026)
- agyn: PhantomFix, a fabricated bug that hijacks an AI autofix agent (Sep 14, 2026)
- Sentry docs: Code Review
- Sentry docs: Autofix
- Sentry docs: Pricing & Billing (Seer pricing)
- Sentry blog: Seer, debug with AI at every stage of development (Jan 27, 2026)
- Sentry docs: Seer
- Visualsoft: Exploring AI issue resolution with Sentry Seer (Jon Purvis, Jul 10, 2025)
- DEV Community: We Almost Deployed an AI Debugging Agent. Then We Found a Critical CVE. (Devji, Dwayo AI, Oct 20, 2026)