A production server is throwing the same null pointer for the fourth time today. In the old world, someone would notice tomorrow morning, after a customer noticed first. In Rollbar's world, the bug arrived in Slack at 03:14, with a stack trace, a list of the deploys that touched the relevant file, and a quiet little badge that said same as issue #4119. Nobody woke up. That is the trick - and the entire product.
Rollbar sells calm. The market calls it error monitoring; engineers call it the thing that finally let them sleep through a deploy. Founded in 2012 by Brian Rue and Cory Virok, two engineers who had just watched the social-gaming company Lolapps melt into a 6waves acquisition, Rollbar started life as Ratchet.io. It was a small idea wearing the wrong name. The pitch was simple: when your code throws, you should see it before your users do.
That idea was less obvious in 2012 than it sounds today. Logs were a wall of plain text. Stack traces were attachments. Most error tracking was a developer staring at tail -f and hoping. Rue and Virok shipped a small JavaScript snippet and a Python library and persuaded a few startups that error data was a product, not a log file. A year later they had a name change, a rebrand, and customers who refused to ship without them.
The first thing you notice using Rollbar is what it refuses to show you. Most observability tools dump everything - every event, every metric, every spike. Rollbar's machine-learning grouping engine looks at ten thousand exceptions and decides nine thousand nine hundred and ninety of them are the same bug. It then hands you ten things to look at, ranked by impact. The dashboard is mostly white space. The white space is the feature.
Brian Rue, who is still CEO, likes to point out that this restraint cost the company speed early on. Investors wanted broader observability. Customers wanted dashboards with more numbers. Rollbar shipped fewer charts and better grouping. By 2017 Bain Capital was leading a $6M Series A. By March 2020, just as the world was learning the word lockdown, Runa Capital was leading an $11M Series B alongside Long Light, Blossom Street and the existing investors. Eighteen million dollars in three rounds is not the loudest funding story in San Francisco. It is, however, enough.
§ 01 — By the NumbersThe shape of a small company doing useful things
Funding Rounds · USD millions
§ 02 — What it actually doesFrom "something broke" to "here is why"
Error Monitoring
Real-time error tracking across more than forty languages and frameworks. The thing that catches the bug your QA didn't.
Automation-Grade Grouping
Turns ten thousand stack traces into the ten that actually matter. The reason engineers stop muting the channel.
Session Replay
Watch the exact session that produced the error. Reproducible becomes a verb instead of a hope.
Resolve AI Agent
Reviews the code, identifies the broken thing, opens a pull request with a proposed fix. You approve, or you don't.
Root Cause
Traces an error backward across services and projects until the underlying cause shows up. Often somewhere boring.
Rollbar MCP Server
Wires production errors directly into AI coding assistants like GitHub Copilot, so the model can answer questions about real outages.
§ 03 — The FoundersTwo engineers who took the wrong job and made it the right one
Former CTO of social-gaming company Lolapps. Left in March 2012 after the 6waves merger and, instead of taking another VP-engineering job, started writing what would become Rollbar in May.
One of Lolapps' first engineers. Joined Rue on the new project the same year. Helped shape the early SDKs and the grouping engine that quietly became the company's moat.
Why the name "Rollbar"
A rollbar does not stop the crash. It keeps you in the seat. The product name is a quiet thesis: software will fail; the job is to make sure your team is still strapped in when it does. It is also a much better URL than ratchet.io.
§ 04 — A short historyFrom Ratchet.io to AI agents
§ 05 — Customers & CompetitionA polite, well-armed corner of the market
Rollbar's customers are mostly the kind of teams that do not advertise their tooling. Engineering organizations at startups and Fortune 500s alike use Rollbar quietly. Public case studies and customer logos over the years have included names such as Salesforce, Twilio, Affirm, CircleCI and Instacart - companies whose engineers spend enough time in stack traces to know that grouping them is worth paying for.
The competition is not shy. Sentry is the loud cousin in the same family. Datadog and New Relic offer broader observability suites that bundle error monitoring in among logs, traces and metrics. Bugsnag (now SmartBear), Raygun, Honeybadger and Airbrake fill out the long tail. Rollbar's bet has always been depth over breadth: do error monitoring better than anyone, and be willing to integrate with everything else.
The integrations list is its own argument. GitHub for code and deploys. Slack for alerts. Jira for tickets. PagerDuty for on-call. Heroku and Vercel for runtimes. The product has spent a decade getting along with whatever else is already in your stack, which is a curious strategy in an industry that loves to be the one ring.
§ 06 — The AI Pivot That Wasn'tWhat happens when an error tracker meets an agent
It would be a stretch to call the past two years a pivot. Rollbar has been doing machine learning on stack traces for most of its life - the grouping engine has been quietly classifying errors for years. What changed in 2024 and 2025 was who could read the classifications.
The Resolve agent is the visible piece. It looks at an unresolved error, walks the relevant code, and proposes a fix as a pull request. Most of the time the fix is right. Sometimes it is wrong in interesting ways. Either way, the engineer is now reviewing code instead of writing it from a stack trace. That is a real change to the job.
The quieter and probably more important piece is the MCP server. Model Context Protocol is a young open standard for letting AI assistants ask structured questions of external systems. Rollbar wired itself in early. Now an engineer using GitHub Copilot can ask, in plain English, what is breaking in production right now, and the assistant can answer with real data. The error tracker stopped being a tab you visit. It became context.
§ 07 — Culture & TeamThirty-five people, a lot of timezones
Rollbar is small. Thirty-five-ish employees, distributed across timezones, with an engineering-heavy roster. The company has been remote-friendly since long before the word became a corporate badge. Its public stack lists Slack, Notion-shaped collaboration tools, Salesforce, Zendesk - the boring, working software that small B2B SaaS companies actually run on.
The culture statement is short. Brian Rue has said that the job is to "support engineers' standards, not just their speed." Translated: do not ship things that make developers regret installing you. There is no founder mythology of all-night hackathons, no aggressively public manifesto. There is a product that has shipped quarterly improvements for more than a decade.
§ 08 — Watch & ReadWhere the company explains itself
Rollbar Product Tour
A walk through the dashboard, grouping engine and core monitoring workflow.
Watch on YouTube →Brian Rue · on building Rollbar
Interviews and talks featuring co-founder & CEO Brian Rue on the origin and evolution of the product.
Watch on YouTube →Resolve & Session Replay with AI
How Rollbar's AI-assisted workflows turn an alert into a proposed pull request.
Watch on YouTube →§ 09 — The Scene, Returned To03:14, and nobody woke up
Back to the server, throwing its null pointer. The old version of this story has someone paged at three in the morning. They wake up confused, stare at a wall of logs, ssh into the wrong machine, eventually find the line, fix it, push, and discover at breakfast that the same bug fired four hundred more times overnight.
The Rollbar version is dull, and that is the point. The alert fires at 03:14. The grouping engine recognizes the exception as the same one that has been quietly accumulating in issue #4119 - tagged a week ago, never resolved. It does not page anyone. It updates the issue's count, posts a single line to Slack, and goes back to listening. At 09:02 an engineer notices the count, opens the issue, and finds that Resolve has already drafted a pull request. They read it for a minute, change two lines, and approve. The fix ships before lunch.
Nobody woke up. That is the entire pitch. Rollbar sells the absence of a 3 AM phone call. It is a less photogenic product than most San Francisco software, but it is the kind of thing engineers thank you for after they have used it for a year. Which is more or less how a small company with $18.4M in funding ends up sitting inside the on-call rotations of companies many multiples its size, quietly catching what the rest of the stack drops.