Application monitoring provider Sentry today announced the launch of Seer Agent, a new feature that enables developers to investigate and resolve production problems using natural language.

Seer Agent uses Sentry’s complete telemetry to surface answers and connections that teams have a hard time finding on their own, significantly reducing the time spent debugging applications.

According to Sentry, developers can ask Seer Agent questions like:

  • Why is this page slow?

  • What caused this spike?

  • What changed before this started?

 While AI coding tools have made it faster to write code, they’ve also created a new challenge: developers are increasingly disconnected from the code itself. When it’s no longer someone on your team placing each line of code, bugs and incidents inherently become harder to investigate and diagnose. Seer Agent makes it easier to cut through the noise to understand exactly what is going wrong, why, and how to fix it.

“When something breaks in production, you’re working across errors, spans, logs, metrics, and more, simultaneously. The volume of data alone makes it hard to know where to start,” Indragie Karunaratne, Sentry Senior Director of Engineering, said in the announcement. “Seer Agent queries all of the sources, connects the relevant signals, and identifies what went wrong and where. Investigations that used to take hours now take minutes.”

Seer Agent is built on three core capabilities:

  • Natural language queries: Ask any question about your application without needing to know exactly where to look in Sentry.

  • Connected context: Surface relationships across errors, spans, logs, traces, and code context that a developer might never have found through manual navigation.

  • Agentic investigation: Walks developers through complex production problems by reasoning through evidence in real time, surfacing what matters from Sentry’s vast data.

“Most teams don’t struggle to know something’s broken. They struggle to know what to fix. Sentry has spent more than a decade building the production telemetry that answers that, and Seer is how we put it to work everywhere developers already are – powering the most complete root cause analysis, automations that hand fixes off to coding agents like Cursor and Claude Code, and opening up our data through MCP and the CLI,” said Milin Desai, CEO of Sentry, in the announcement. “In Slack, the investigation becomes multiplayer. The dev team can swarm an incident, redirect Seer mid-step, and leave the thread behind as a record of how it got solved. Seer Agent is one more way to engage with it.”

Seer Agent is now available in Slack, allowing users to start an investigation by messaging Seer Agent in any channel. It makes the experience multi-player, by allowing anyone in the channel to query it, redirect mid-step, and add context the agent didn’t previously have. Alternatively, channel participants can watch the team go from incident to resolution as an observer to better learn the system.

More information is available here.