Hardware-in-the-loop testing for engineers and their AI agents.

Stout is a control plane for hardware-in-the-loop testing. It runs hardware tests on your pull requests, sorts the failures into firmware bugs, bench problems, and flakes, and manages the fleet of Lager test boxes the tests run on. Under it all is the open-source Lager runtime.

When a pull request opens or updates, the reviewer reads the diff, plans a hardware test with tool access to the Lager runtime, picks a box, runs the test, and posts pass or fail on the PR. By default a person approves the plan before it runs. You bring your own AI provider key, either Anthropic or any OpenAI-compatible endpoint.

CI Metrics tells you what your hardware CI is really finding. It reports which merges were checked on hardware, which failures were real firmware bugs, which were caused by a bench or harness change, and which tests are flaky. An admin can confirm or dismiss any row.

Yes. Agents open pull requests like anyone else, and the PR Reviewer tests them the same way, on the same hardware. Nothing in your agent setup has to change.

Lager is an open-source (Apache 2.0) hardware test automation platform. A Lager box is a Linux machine that runs the Lager runtime in Docker and drives the instruments on a test bench, such as oscilloscopes, power supplies, debuggers, and the devices under test, through a Python API and CLI. Stout manages fleets of these boxes.

No. Lager is the open-source runtime that controls your hardware; Stout is the control plane that governs and scales it across teams. It connects to your existing Lager boxes unchanged, and Lager stays Apache 2.0, so if you ever stop using Stout, your boxes, scripts, and tests keep running on vanilla Lager.

Generic CI runners assume interchangeable compute. A test bench is the opposite: each box has specific instruments wired to a specific board, and two jobs cannot share a power supply. Stout adds the hardware layer. It picks a box, locks it, runs the test, streams the log, and posts the result to GitHub as a check. It plugs into the CI you already have instead of replacing it.

Yes. A PR Reviewer runs on every pull request through the Stout GitHub App and posts a check run, so branch protection can require a hardware pass before merge. Your own CI can also report its hardware runs to Stout, and CI Metrics shows which merges were checked on real hardware.

Users sign in through your identity provider over SSO (OIDC or SAML), with just-in-time provisioning on first login. Role-based access control scopes exactly who can view and run what across organizations, teams, and individual boxes, and removing a user in your IdP immediately revokes their access.

Every action is recorded in an immutable audit log (who did what, on which box, and when), and accounts can require MFA. Stout runs as a managed cloud service: your boxes open an authenticated outbound connection to it, so there are no inbound ports to expose on your test network.

Pricing scales with your team and fleet size. Book a demo and we'll put together a quote that fits.