Operations | Monitoring | ITSM | DevOps | Cloud

Join Serge for an introduction to tracing and how it applies to AI agents. Through live walkthroughs, you’ll see how to follow an agent’s execution, inspect model and tool calls, and find where things went wrong. No prior tracing experience is needed.

We use Sentry's new Regex search for Logs to hunt for non-obvious bugs in our app. What do you do when traces show that Postgres jobs are backing up? Logs are trace-connected, so by looking at one of the affected traces, we can see all of the logs on that trace. In there, we can see Postgres has logged that it acquired a lock only after a long wait. Using regex, we can find all of these logs that describe acquiring a lock after a 10-second-plus wait. That narrows our search from thousands or hundreds of logs down to about a dozen.

This video walks through what it looks like to treat Sentry's Seer agent the way you already treat a chatbot, asking it plain language questions about your own production application. From Slack or inside your Sentry account, Seer Agent can answer questions with real context: the affected spans, the recent release, the likely cause. Seer doesn't stop at answering questions. Ask it to act on what it just told you and it completes that work inside Sentry, serving as an assistant to your projects, team workflows, and your future self.

Welcome to HotFix, the show where Sentry customers return the favor. We helped Exa fix their code. Now, it's their turn to fix our organization.

Checkout suddenly got a lot slower for a chunk of users, and the only clue was a spike in an application metric. In this walkthrough, we go from that spike to a linked trace that shows exactly where the time went. Using Codex alongside Sentry, we dig into the trace data, isolate the slowdown to the European backend, and follow it down to a single Stripe request that's dragging everyone down with it. Along the way, we talk through what actually helps here: caching, timeouts, async calls, and other ways to keep one slow dependency from becoming everyone's problem.

As fast as AI can generate code, getting that code to production still tends to slow things down. It makes sense. When code breaks in production it’s not just a nuisance; it’s a full-blown incident. And can have big impacts for your company and customers. The question really is, can you catch broken code as quickly as you can ship it? In this livestream, Peter McCarron, Technical Product Marketing at Sentry, is joined by Tom Totenberg, Head of Release Automation at LaunchDarkly, to talk about how you can automatically protect new releases from code merge to error detection.

Welcome to HotFix, the show where Sentry customers return the favor. We helped Golden State fix their code. Now, it's their turn to fix our CEO’s jump shot.

Agent Tracing shows the full execution path of an AI agent: the model call, every tool invocation and its arguments, token counts, cost, and the span where it broke. Same traces and spans you already use, with agent-specific attributes on top. Serge walks through three apps — a Next.js e-commerce agent using the AI SDK with a failing tool call, a Slack bot built with Eve that orders lunch, and a code review agent built with Flue over MCP.

Tracing does a lot more than help you monitor performance. It's what ties errors, logs, and session replay together in Sentry, so you, and your agents actually know what happened. It's included in every Sentry plan.

Sentry brings errors, logs, replays, spans, and metrics together to find and fix the worst issues in your application.