13.08.2025 18:45
👍 0
🔁 0
💬 0
📌 0
Engineers should be running autonomous coding agents in the cloud. Give the agent its own resources, and let it crank away for a long period of time without a human having to review or intervene. Here's an example of running Claude Code on RWX. Full details in the blog post ↓
ReadWriteExecute, Inc. announced that it completed a System and Organization Controls (SOC) 2® Type 2 Audit examination for their System in August 2025. The company has been SOC 2 Type 2 compliant since 2023.
www.rwx.com/blog/soc-2-t...
We shipped an MCP Server via the RWX CLI. Use it to have AI fix tests that failed on CI, powered by RWX tracking test failures as a semantic output.
More details in the blog post below ↓
To accelerate performance in high iops scenarios like an npm install, we shipped support for running tasks using tmpfs.
In July we shipped ARM runners, a VS Code Extension, and an option to start runs manually. Stay tuned for progress on a new way to build container images. See the full recap below 👇
www.rwx.com/blog/july-20...
The GitHub Actions base image is a whopping 47 GiB. It includes a list of bundled software a mile long.
A bloated image affects performance. But the biggest problem is that it makes build scripts far less portable.
www.rwx.com/blog/github-...
BuildKit was released in 2018 and made several improvements over the classic Docker builder, but it still leaves a lot to be desired.
This is our proposal for a new way to build container images resulting in substantially faster builds with simplified configuration.
www.rwx.com/blog/proposa...
RWX is the only CI/CD platform that automatically caches every task. Cache keys are determined automatically, and tasks are sandboxed so that they can only access the files specified in the filter. Sandboxing is the key to ensuring there are never false positives on cache hits.
It's really insane on GitHub Actions that when a single job fails, it cannot be retried without canceling the entire workflow. On RWX, tasks that fail can be retried immediately, even while other tasks are still running.
We just shipped support for running arm64. Building arm containers is 10x faster without emulation. We built the initial RWX image for arm by running on x86 with qemu, and it took 14m 8s to build the image. Building arm on arm took 1m 38s.
We just shipped an API endpoint to set vault secrets. Programmatically setting secrets is especially useful for automating secret rotation.
We just shipped the ability to manually initiate CI runs on RWX. Engineering teams can choose to make CI opt-in to push code freely without overly running CI and incurring higher costs.
We recently shipped support for specifying the operating system used for RWX runs. The CLI injects the default configuration, providing explicit configuration without having to write boilerplate.
We recently shipped an RWX package to create GitHub pull requests. It's the fastest and easiest way to automate opening PRs.
The first run in the demo video had a typo in the config. Fixing it took seconds rather than minutes due to automatic content-based caching.
We're excited to welcome Dave Reed to RWX as our Chief Operating Officer.
www.rwx.com/blog/dave-re...
We shipped a web-based file tree navigator to see files from a task's output. To make browsing as fast as possible, we implemented it in WebAssembly.
We recently added the capability to download logs for a group of tasks. At first, we thought we were going to have to build an asynchronous user experience commonly used for batch downloads. RWX is a platform built on performance though, so we made downloads start immediately.
