CI that caches the whole machine.
Every green run on main is saved as a snapshot: code, dependencies, build caches, Docker layers. Each job restores its own copy in seconds.
The CI of 10 open-source projects · 2 vCPU machines
6×faster
than one job on a GitHub-hosted runner.
Time to a green check
5×faster
than one job on a Blacksmith runner.
Time to a green check
7×faster
than one task on Vercel CI.
Time to a green check
Medians across Django, React, Prettier and seven more, timed from the push to a green check. See the benchmarks →
Connect GitHub and add a config
Install the boxd GitHub App and list your checks in one small config. GitHub triggers the run and shows the result. boxd runs it.
Every push and PR runs on boxd
Fresh machines are restored from your snapshot, run the checks and test shards at once, and post a GitHub check.
The cache updates itself
When a run on main passes, its machine is saved as the new snapshot. The next job starts from there.
Also in every machine
GitHub checks
Results on every commit and pull request, with the time of each step.
Configured outside GitHub
No workflow files: the config lives with the automation on boxd.
Isolated machines
Code under test gets no tokens and no access to your account.
Secrets
Sealed env vars for the jobs that need them.
Egress allowlist
Jobs only reach hosts you name.
Bigger machines
Up to 8 vCPU / 32 GiB per job, billed per second.
SDKs
Script your own CI flow in Python or TypeScript.
Self-host
Run boxd in your own cloud or datacenter.