Meet Sera, your AI SRE teammate

Think about the best person you've ever worked an incident with.

They didn't take over. They didn't need a briefing. They showed up already knowing what was going on, pointed you at the thing that mattered, kept everyone else off your back, and wrote it up afterwards without being asked. You still made the calls. You just made them faster, and you weren't alone.

That's the bar we set for Sera.

A teammate, not a tool

Most AI in operations is built as a tool: you point it at something and it does one job. That's useful, but it isn't how incidents work. An incident is a chain of moments, each one needing something different, all of them under pressure.

A teammate carries context from one moment to the next. It remembers what it saw when the alert fired when you ask it a question an hour later. That's the difference we care about, and it's why Sera is one teammate rather than five separate features.

How Sera works the incident with you

Sera Summary. The moment an incident opens, Sera tells you what's going on in plain English: what's failing, which service, who's affected. You start with the picture instead of a raw alert payload.

Sera Investigate. This is where the time usually goes, and where Sera does its most important work. It reads the alert, correlates related alerts into one incident, searches your past incidents for similar ones, and checks recent code changes in GitHub. You get a probable cause, a confidence level, the evidence behind it, mitigation steps in order of urgency, and a link to the exact pull request and line that likely caused it.

Sera Scribe. When the war room forms, Sera joins the Zoom or Teams call, writes a live timeline and keeps stakeholders updated. Nobody pulls you off the fix to ask for a status.

Sera Querier. Mid-incident, you need a specific answer. @mention Sera in Slack with a plain-language question, like "what's the threshold on this alert?", and get the exact answer from the incident data.

Sera Postmortems. When it's resolved, Sera drafts the postmortem from everything it already gathered: the timeline, the call, the investigation. You review and edit. You don't start from a blank page at the end of a long day.

What stays with you

We made one decision early and haven't moved on it: Sera doesn't make changes to production on its own.

When something goes wrong in production, a human name is on it. Engineers know that, and they're right to be careful about any tool that wants to act without asking. So Sera suggests, and you decide. It shows its confidence and its evidence for every conclusion, so you can check its reasoning in seconds. When it isn't sure, it tells you.

Where Sera works

Sera lives inside Xurrent IMR, and it meets you where you already are: in the IMR app, in Slack, on your incident calls, and inside Claude through MCP. You don't need a new tab to live in during an incident.

Why we built it this way

Engineering got a lot of AI this year. Most of it went to writing code faster. The people who keep that code running got very little.

We think that's backwards. Reliability is the hardest job in software, and the people who do it deserve help that respects how they work: help that knows their systems, shows its reasoning, and leaves the decisions to them.

That's Sera. Your AI SRE teammate.

‍