Nebula

Keep builds moving without paying for idle runners

Nebula gives each GitHub Actions job a dedicated machine, then shuts it down when the build finishes.

Match compute to the build

Choose the capacity your build needs.

Choose your preferred hardware

Run builds on supported ARM or Intel machines.

Pay less for build compute

Use discounted AWS capacity, with standard-priced capacity as a fallback.

Keep each build separate

A runner handles one job, then shuts down.

CAPACITY WHEN NEEDED

Get the right capacity when a build starts

Nebula starts a matching machine when GitHub queues the job, then shuts it down when the build finishes.

  • Choose the size you needSelect from available machine sizes for each workload.
  • Avoid idle runnersStop paying for the runner when its job is done.
  • Keep access short-livedEach runner receives access for its assigned job.
FASTER BUILDS

Download less, build sooner

Reuse saved dependencies instead of downloading them for every build.

  • Keep data close to the runnerUse a cache near the machine running your build.
  • Reuse saved dependenciesRestore cached files rather than download them again.
  • Avoid repeat downloadsKeep shared dependencies available for the next job.
COMPUTE OPTIONS

Choose a machine for the job

Match capacity to the build, from quick checks to large test suites.

CPU-HEAVY

Compute optimized

Use for compilation, type checks, and CPU-heavy test suites.

GENERAL

General purpose

Use for integration tests, browser tests, and service matrices.

MEMORY-HEAVY

Memory optimized

Use for large repositories, database tests, and memory-heavy suites.

From queued build to completed work

One annual company license, with compute billed separately by AWS.

  1. 01

    Choose the build capacity

    Set the machine size and type for the job.

  2. 02

    Start on demand

    Nebula starts the selected machine when the job is ready.

  3. 03

    Run, then shut down

    The build finishes and the machine shuts down.

Self-hosted GitHub Actions runners

One flat license. Compute at cost.AWS bills you for the instances at spot prices, with no markup. You pay us one license a year per company, sized by the jobs you run: no seats, no minute-metering.

150,000 runner minutes a month

Compare GitHub-hosted Linux 8-core arm64 with self-hosted runners on the same workload

Estimate for 8-core arm64 in Virginia, US. Self-hosted runners total includes the annual license at its monthly equivalent; AWS storage and networking are excluded. Spot pricing varies (quote dated 7 October 2026).

$240/ year

Billed once a year