How Warp runs everything from CI to Oz, their cloud agent platform, on Namespace
“Namespace is the best dev infrastructure tool I’ve used in years. We run everything from our CI to Oz, our cloud agent platform, on Namespace, and I could not recommend it enough.”
Aloke DesaiHead of Product Engineering, WarpWarp is a cloud agentic development platform with three core products: a modern terminal, a built-in coding agent, and Oz, a cloud agent platform that lets developers deploy and run cloud agents at scale. They are trusted by over 800,000 developers and thousands of engineering teams at leading companies like Ramp, Amazon, GitHub, Docker, VMWare, and so many more.
When Warp needed fast, reliable infrastructure for Oz, Aloke turned to a platform he already knew well. His team had been running CI/CD on Namespace for months and was seeing significantly faster builds than with their previous setup, so they saw a chance to use that same foundation as the backbone of their cloud agent platform. But the story starts well before Oz, with a fleet of intern laptops in a New York City office running GitHub Actions.
Problem 1: A fleet of intern laptops running GitHub Actions in a NYC office
Warp’s terminal is built entirely on Mac, and historically their CI ran on GitHub Actions. Mac runners on GitHub are notoriously slow and expensive, and Aloke’s team quickly hit a wall: CI builds were taking up to an hour per run. To work around it, they repurposed old intern laptops as self-hosted runners (yes, really). But the fix created its own problems.
“We actually ended up using old intern laptops to build self-hosted runners. But it wasn’t really maintainable. They were only in New York, so if someone wasn’t in New York and they had to maintain them, that wasn’t an option.”
As Warp grew and prepared to open source their terminal, running CI on intern laptops wasn’t going to cut it.
Solution 1: Selecting Namespace over GitHub and Self-hosting
Coming off a fleet of intern laptops, Warp set a simple bar: faster Mac builds, zero new operational burden, and a migration they could do in an afternoon. Namespace checked every box. Under the hood, our platform offers macOS instances running on Apple M4 Pro and M5 Max processors for native Apple Silicon performance, and our GitHub Actions runners are directly built on top of that same platform. So when Warp's team pointed a handful of Mac builds at Namespace, they immediately saw 4x faster builds. And because the runners are a drop-in replacement for GitHub-hosted runners, the migration really did happen in an afternoon.
When it came time to open source their terminal, moving everything over was the obvious next step. By the time it was fully open sourced, 100% of CI was running on Namespace. Today, every PR from every contributor now builds entirely on our infrastructure.
“Our CI before and after Namespace was night and day. My only regret is that we didn’t migrate sooner.”
Problem 2: Scaling a new cloud agent platform
Five months after Warp first moved their CI/CD onto Namespace, Aloke began thinking about Oz, the upcoming cloud agent platform that would need reliable, sandboxed compute to run agents. As with most ambitious infrastructure projects, though, it’s simpler in theory than in practice.
Aloke found himself asking a lot of hard engineering questions: how they would keep build times under control, how they would safely run thousands of long‑lived, stateful cloud agents for different enterprises, on different operating systems. As those questions and more piled up, it became clear he needed a purpose-built platform rather than another ad hoc workaround.
Solution 2: Selecting Namespace as the backbone of Oz’s agent infrastructure
When Aloke’s team started looking for a purpose-built platform for building Oz, they already knew who to call. However, this time the requirements looked different from CI/CD. Oz needed strong isolation between enterprise tenants, configurable sandbox restrictions, and OS-agnostic support including Mac and Windows. On top of that, it also needed flexibility around machine shapes, the CPU, memory, and storage resources allocated to each compute instance.
In the end, Namespace’s platform felt like the natural fit, not just for the feature set, but for the track record. Namespace had carried them through the entire CI migration without any reliability hiccups, and by that point Aloke’s team saw it as infrastructure they could bet Oz on.
“It was a combination of seeing success with CI/CD and knowing the Namespace team was there to help support us and work deeply together to see positive outcomes.”
Results: 4x faster GitHub Actions and 20k sandboxes per month
Warp’s CI builds went from roughly an hour to 12 to 15 minutes, a 4x improvement, with a one-line migration. Hundreds of builds and integration tests run on every PR without the team thinking about the underlying infrastructure.
“Any time you ask a question in Slack, they go above and beyond to respond really, really fast. And not just necessarily fix the problem, but go a level deeper to fix the platform. I think that’s what causes us to be very invested in Namespace.”
For Oz, Namespace now powers every cloud agent run. The platform provides the isolation, flexibility, and observability that Warp’s customers expect. When runs fail, Aloke’s team digs into our dashboard to find root causes quickly, whether that’s an OOM event or a configuration issue, rather than guessing.
“Namespace is a core piece of our agent infrastructure right now. It has to work. And I found it very reliable.”
If you’re building cloud agent infrastructure or your CI is just slow, give us a try. It just works.