Capstanworks rebuilds the way your team ships, until a release is something nobody has to watch.
A cluster your team can actually operate, rather than one only its author understands.
Pipelines that finish in minutes and tell you plainly which line broke them.
Dashboards somebody opens on a normal Tuesday, not only the ones opened during an outage.
Moved in slices you can stop between, with a way back from each one until it holds.
The bill read line by line, then brought down without anybody being told to use less.
What we work in
Principal Engineer
A team that ships once a fortnight is not being careful. It is being slowed by something nobody has measured.
Capstanworks was started in 2017 by four platform engineers who had each spent a year somewhere watching good teams lose their afternoons to a build. We work with companies that already have engineers and already have a cloud bill, and we leave behind a pipeline, a cluster and a set of dashboards the people who stay can run without us. Every engagement ends with a handover week, and we mean the handover, not a document.
Two days a week on your build and deploy path. We take the pipeline, keep it under ten minutes as the codebase grows, and answer for it when it breaks.
Your cluster, your infrastructure code and your dashboards, run with you rather than for you. Two engineers, in your standups, with your team doing the work by the second month.
A move to or between clouds, cut into slices you can stop between. Each slice has a rollback, a cost estimate before it starts and a number afterwards saying what it actually came to.
Three of our engineers inside your team for two quarters, working your board and your on-call. For a platform team you are still hiring into and cannot wait to finish.
Before
Release night, once a fortnight.
Four people on a call at nine in the evening. A checklist in a document somebody wrote three years ago, a build that takes fifty minutes and fails twice for reasons nobody writes down, and a rollback plan that has never been rehearsed. The change itself took a day. Getting it out took the week.
After
Tuesday afternoon, nobody watching.
The same change merges at two and is live by half past, because the pipeline is eight minutes and the deploy is its last step. If it misbehaves the rollback is one command that has been run in anger before. Nobody stays late, nobody is on a call, and a release stops being an event anyone has feelings about.
Not sure which retainer fits? Tell us how you release today and we will say which it is.Talk to an engineer
Prefer email? Write to us directly. hello@example.com