Seer

Seer is the AI debugging agent built into Sentry, the error and performance monitoring platform. It reasons over the telemetry Sentry already collects from your running application, including errors, traces, logs, replays, and profiles, to find what broke and why and then propose a fix. According to the Sentry homepage, Seer can investigate a new issue automatically so that a pull request is waiting before an engineer even opens their laptop.

What It Is

Seer is Sentry's AI layer for troubleshooting and fixing errors and performance issues. The docs describe it as an end-to-end debugging agent that works across the full context of your application and often catches problems before they are merged. It is turned on per organization after the Sentry SDK is installed and a GitHub repository is connected.

How Autofix Works

Autofix automatically analyzes issues as Sentry ingests them. According to the docs, it runs a root cause analysis step that decides whether an issue can be fixed with a code change and adds an initial guess about the problem to the issue page. From there Seer can generate a code fix and open a pull request on GitHub, and an Issue Scan feature triages which issues are most likely fixable. At the final code generation step, you can hand the implementation off to an external coding agent instead, with documented support for Claude Code and Cursor Cloud Agents.

Asking Questions with Seer Agent

Seer Agent, which the docs note is in open beta, lets you ask questions about your application in plain language and pulls the relevant telemetry to answer them. The docs say it connects evidence across errors, spans, logs, traces, and code context, reasoning through production problems in real time. Conversations can be shared with teammates, copied to other agents, or revisited later.

Catching Bugs with Code Review

Seer also includes AI Code Review, which reviews code changes on GitHub pull requests, predicts errors, and suggests improvements before a merge. The docs state this feature is available only on GitHub.

What It Runs On

Seer runs on what Sentry calls its context layer. The docs list the data it can examine, including issue context such as stack traces and event metadata, distributed tracing data, structured logs, relevant code from linked GitHub repositories, performance profiles, and Sentry's own documentation. Sentry says Seer supports all of the programming languages and frameworks that Sentry supports across web, mobile, and desktop.

Debugging Where You Work

Seer is not limited to the Sentry interface. The product pages describe fixing bugs from a Slack alert, starting an incident investigation in Slack, and sending fixes to tools like Claude, GitHub Copilot, or Cursor, with the Sentry MCP server connecting issues, traces, and logs into external AI agents and editors.