Getting started

hypervise is in private beta. Install the app, and we enable your installation (usually within a day: write to hello@hypervise.sh).

1. Install the GitHub App

Install hypervise on the repositories that should run on it (Install button on the home page). It asks for:

• Repository administration (write): to register a one-job runner for each job. That is the only use.
• Actions (read): to read workflow job metadata.

2. Use the hypervise runner

jobs:
  test:
    runs-on: hypervise
    steps:
      - uses: actions/checkout@v4
      - run: make test

That's all. The job runs in a hypervise VM: Ubuntu 24.04 with the GitHub runner, Docker, build tools, Python, Rust, Go and Node available through the usual setup-* actions. It is recorded from the first step, on one CPU: safe, but slower than a normal runner.

If the job fails, its recording is kept: open the job on your dashboard to see it, and to replay it.

3. Optional: make it fast where it doesn't need recording

    steps:
      - run: hypervise update-config --physical-cpus 4 --no-recording
      - uses: actions/checkout@v4
      - run: cargo build --workspace --tests && cargo fetch
      - run: hypervise update-config --shards 4 --physical-cpus 1 --recording
      - run: cargo nextest run --partition count:$((HV_SHARD_INDEX+1))/$HV_SHARD_COUNT

• Checkout and build run unrecorded on 4 cores.
• The tests run as 4 recorded shards, a quarter each.
• Each shard's output appears in the step's log, in its own group.
• A failing shard is kept as a recording: see it on your dashboard.

Next: the workflow model.