Back to blog
Three nested teal glass chambers with a luminous mint centre.
Launch

Introducing Borealis: a place for your code to run

Borealis gives developers and their agents managed sandboxes for running tools, changing files, and starting services. Here is why we built it.

Borealis Team

Borealis went live on 21 July 2026 as a managed home for sandbox workloads. A sandbox has its own image, resources, identity and lifecycle, so whatever runs there can install dependencies, inspect a repository, change files, run tests and start a service, with each step finding the state the previous one left behind.

Give the work an environment

A coding task rarely ends with one command. Something installs dependencies, inspects a repository, changes files, runs tests, and starts a service to check the result, and each step depends on the state left by the previous one. Whether a person types those commands or an agent issues them, the environment has the same job.

Borealis makes the sandbox an explicit part of that workflow. Choose a container image, set the resources and startup configuration, and give the task an environment in which to work. The sandbox has its own identity and lifecycle, so you can inspect it and decide when it should run or stop.

That matters as much for the person watching as for the agent working. When a tool fails, you need access to the environment and its output to understand why.

Keep control of execution

Different jobs need different amounts of compute. A small script, a dependency install, and a development server should not all require the same environment configuration.

Borealis lets you choose the image and resource allocation for the work. Commands and terminals give you a way to run tools and inspect their results. Services running inside a sandbox can be made reachable through configured ports, so you can check what the workload has produced.

The goal is a practical working environment that you can observe and manage throughout a task.

Build around the workload

Borealis focuses on execution infrastructure. You choose the application, agent, and tools that use it. Your application remains responsible for deciding what work to request and how to interpret the result.

The judgement calls belong in the workflow you build around it: whether a command is appropriate, whether an application is healthy, whether a change is ready to ship.

Start with one sandbox

Choose a small task: run a script, try a tool, or start a development service. Create an environment for it, inspect the result, and stop the sandbox when the work is done.

The getting-started guide walks through that first session. The documentation covers the product as it works today, including capabilities added since launch. We will use this blog to explain those additions and the engineering behind them.

Create a Borealis account or sign in to get started.


Launch retrospective, written on 30 August 2026 and filed under the original go-live date. Linked documentation reflects the current product.