Built for the agent loop
Coding agents finish a change in under a minute. Then they wait twelve for the test suite. Offload closes that gap by running your existing tests across isolated cloud sandboxes, in parallel, with no code changes.
Parallel by default
Offload spreads your suite across 200+ isolated Modal sandboxes. Each test gets its own container. LPT scheduling packs the batches so the slowest test sets your total runtime, not your CPU count.
Drop-in for your framework
All it takes is one TOML config dropped into your repo to support pytest, cargo-nextest and vitest out of the box.
Flakiness has a fix
By isolating test runs to sandboxes, we avoid things like port collisions and tempfile races that can run local parallel runs. Genuinely flaky tests get retried, so one bad result doesn’t force you to rerun your entire suite.
Features
Run tests faster with less local resources
Offload runs your tests on Modal sandboxes, which start up in a jiffy. We orchestrate the distribution, monitoring and gathering of results. You can customize offload to retrieve test artifacts as well.
Keep your existing tests
Offload reads your existing test suite. This means there’s no framework migration, no new test syntax to learn, and only one simple config file to write.
Enhance test reliability with isolation
Nothing is more frustrating than a suite which flickers from green to red in CI. A whole class of flakiness is eradicated by isolating the environment your tests run in sandboxes.
Built by people who care, for people who build
We built Offload because we could see that because agents could rapidly write code and tests, we’d naturally end up with lots of tests. Running those tests requires time and resources, drawing both away from what’s available to our agents. And this problem multiplies the more agents we run in parallel.
Offload solves a real problem that we have because we built it to fit the new way we work now. Because Imbumans run many different kinds of projects, Offload was designed to be small, simple and flexible. These attributes mean that Offload can be adopted in your own project with a minimum of fuss!
Frequently asked questions
What is Offload?
Offload is an open-source Rust CLI that runs your existing test suite across 200+ isolated cloud sandboxes in parallel, instead of on your local machine.
Do I need to rewrite my tests to use it?
No. Offload discovers tests through your existing framework (pytest, cargo-nextest, or vitest) and runs your existing commands. You write one TOML config and check it into your repo.
Which test frameworks and providers does it support?
Pytest, cargo-nextest, and vitest are supported out of the box, with a default framework type for any custom runner. Sandboxes run on Modal today, and the provider interface is designed to support other bash-scriptable infrastructure.
How much does it cost to run?
Cost depends on your sandbox count and test duration. Our 345-test Playwright suite runs for about $0.08 on Modal, billed to the second.
What happens when a test is flaky?
Offload retries failed tests up to your configured retry_count, in parallel, across available sandboxes. A test that passes on retry is marked flaky, not failed, so a single flake never fails your whole run.
Is Offload free and open source?
Yes. Offload is MIT licensed and available on GitHub. Install it with cargo install offload.