Depot

Depot is CI/CD and build infrastructure for engineering teams: a CI engine that runs GitHub Actions workflows, hosted runners, container builds and a cache.

Depot shortens the wait between a pushed commit and a shippable build by replacing the runners, builders and caches behind a team's pipeline. Depot Technologies Inc. builds it and runs every part on its own compute and orchestration layer. It is designed for a shift in how code arrives, with engineers directing AI agents that open changes in parallel. A team can adopt one part to clear one slow stage, or move its whole delivery pipeline onto Depot. Depot CI is a programmable CI engine that runs GitHub Actions workflows stored in the repository's .depot/workflows folder. It handles matrix strategies, reusable workflows, cron schedules, OIDC credentials for AWS, GCP and Azure, and secrets masked in every log. Depot GitHub Actions Runners swap GitHub-hosted machines for Linux, Windows and macOS runners, Arm included, with analytics on slow jobs and failure rates. Container Builds produce Docker images on native Intel and Arm builders, Depot Cache keeps one persistent cache shared across builds, and Depot Registry stores the images. Depot Code adds Git-based source control, while Depot Sandboxes and Depot Managed Agents bring the same compute to AI coding agents. Depot fits engineering teams already on GitHub Actions whose workflows queue on underpowered runners or wait on slow image builds. Platform engineers who maintain their own runner fleets on AWS can hand that work over; PlanetScale replaced its self-hosted Buildkite setup with Depot. Teams running many AI coding agents get isolated sandboxes and a cache every agent shares. Companies with strict data rules can place the Depot data plane, its build compute and cache, inside their own AWS account through Depot Managed. A team can point one slow job at Depot, such as a multi-platform Docker build, and leave the rest of its pipeline alone. Container Builds plug into existing pipelines on Buildkite, CircleCI and Bitbucket Pipelines as well as GitHub Actions. Moving a whole workflow across starts with the depot ci migrate command. Its test results then mark any test that both passes and fails on the same commit as possibly flaky. When a job breaks, an engineer can SSH into it while it runs, inspect CPU and memory metrics and search logs across runs instead of guessing from a rerun.