How Sentry Uses Claude Managed Agents to Auto-Fix Bugs

When AI already knows exactly where the bug is, but a developer still has to fix it by hand, that gap is the real bottleneck. Sentry just closed it with Claude Managed Agents.

The real problem at Sentry

Sentry processes billions of events every day and already had an AI agent called Seer that pinpoints root causes with remarkable accuracy. The catch was that developers still had to write the fix themselves and open a pull request. The distance between "here is what's broken" and "it's fixed" was draining the team's momentum. As Sentry put it: "Telling a developer where the bug is isn't enough anymore, they want you to actually fix it."

A complete, automated pipeline

Sentry plugged Claude Managed Agents in right after Seer, turning the whole thing into an end to end pipeline:

• Seer queries the source code alongside telemetry data (stack traces, profiling, logs, spans, metrics)
• Seer builds a comprehensive Root Cause Analysis
• That analysis gets handed off to a Claude Managed Agent
• The agent plans the fix, implements the code change, and opens a pull request automatically
• The developer only needs to review and approve it

From the moment a bug is detected to having a ready to merge PR, the entire process now runs on its own.

The numbers that get everyone's attention

Here's what the rollout actually produced:

• Over 1 million Root Cause Analyses per year
• More than 600,000 pull requests reviewed every month
• Just one engineer was needed to build the entire integration
• Shipped in a few weeks instead of the months it would have taken to build the infrastructure in house
• Completely eliminated the cost of running and maintaining separate agent infrastructure

No compromise on security

Sentry accesses Claude through Vertex AI specifically to keep data residency inside Google Cloud. No new subprocessor to vet, no added compliance headaches. That matters a lot for companies handling sensitive data at scale.

What developers everywhere can take from this

Indragie Karunaratne, Senior Director of Engineering for AI/ML at Sentry, put it this way: "Managed Agents didn't just let us ship this integration in weeks instead of months, it also completely removed the burden of running our own agent infrastructure."

The takeaway is worth sitting with: you don't have to build everything from scratch. With managed agent infrastructure, even a small team can roll out automation at enterprise scale. What AI agent are you already using in your daily workflow?