RWX raises $12M as the engineering bottleneck moves from writing code to validating it

August 4, 2026·
Dan Manges
Dan Manges (Co-founder, CEO)

In the past year, the way software is built has radically changed. The octocat has served us well for the past two decades, but it's time to rebuild the software forge for agentic software engineering.

With engineering organizations steadily migrating from GitHub Actions to RWX, we're excited to announce that we've raised $12M in Series A financing led by Hyde Park Venture Partners, with Quiet, The OHIO Fund, R1, and DV participating, along with angels including Kohsuke Kawaguchi (creator of Jenkins), Bryan Johnson (Don't Die), and Theo Browne (t3.gg).

The best engineering teams are re-evaluating every dev process and part of their stack to adapt to building software with agents. We built RWX from first principles to accelerate the process of building, testing, and deploying software. The approach that we took to improve the developer experience for engineers positioned us to work exceptionally well with coding agents.

CLIs are the de-facto interface for agents, and code execution on RWX is fully decoupled from version control webhooks. Agents can test code in sandboxes, run full CI pipelines, and deploy preview web apps, all without a single git commit or push.

Code execution on RWX is powered by an entirely new container runtime. RWX is compatible with the OCI container specification, but it runs much faster due to state-of-the-art techniques like automatic content-based caching, graph-based task execution, and cloud-native optimizations.

Why we started with CI/CD

We built RWX from experience leading high-growth startups. I was the founding CTO of Braintree, which acquired Venmo and was then acquired by PayPal in 2013 for $800M. A couple of years later I cofounded Root, which launched in 2016 and went public via an IPO in 2020 with a $6B market cap.

As an engineer focused on shipping, I hate waiting. I've spent countless hours tweaking my vim config to shave milliseconds off workflows. Even worse than waiting is doing so unnecessarily. Without RWX, CI pipelines have large amounts of redundancy, running the exact same commands on the exact same code ad nauseam. An engineer makes a change to one small portion of a codebase, and a massive fleet of build infrastructure kicks off, needlessly rerunning the same setup steps as the previous job and verifying that unchanged code is not broken.

Accelerating feedback loops has always been a key to engineering productivity. The faster you can iterate, the faster you can get to the end state and ship.

Although coding agents can operate autonomously, feedback loops are still relevant. The faster an agent can iterate and validate its work, the faster it can fulfill its prompt, and the fewer tokens it'll use. The interfaces and workflows need to change though. Designing for agents means building CLI-first interfaces, infrastructure to support high volumes of autonomous activity, and the right UX to put humans in the loop only when necessary.

When writing code took two hours, a ten-minute CI run was acceptable. When an agent writes the same code in five minutes, CI suddenly consumes 67% of the development cycle.

Jared Palmer, former SVP at GitHub

What people (and agents) are saying

Engineers come to RWX with longstanding frustration towards current solutions in the market. Inefficiencies that were tolerable pre-AI are now unbearable as teams experience a sharper loss of momentum between writing and validating code.

In our case study with Honeycomb, they described what changed after adopting RWX:

Our CI system couldn't keep pace with AI-assisted development. RWX's content-based caching changed how we thought about CI. We could stop rerunning work that hadn't changed and keep validation fast as build volume increased.

Dean Strelau, Staff Engineer at Honeycomb

The speedups are often significant enough that people's first reaction sometimes includes skepticism:

We encountered two-second builds and folks had questions.

David Irvin, Senior Engineer at Coalesce

Even the bots are confused! Last week, one of our engineers shared Claude's reaction after an RWX workflow completed:

29s for three Node versions including a Chrome install is too fast to trust. Checking whether the nested runs actually got my patch.

Claude, Robot at Anthropic

When both engineers and AI agents assume the system must have failed when it did not, it's a sign their expectations were built around the limitations of yesterday's CI.

Getting started

The best engineering teams have always adopted the best engineering tools. If you're overhauling your development workflows for AI-driven engineering, we’d love to help you get started using RWX.

To get started on your own:

Pair RWX with the coding agent you already use, and let it migrate your build for you.

 brew install rwx-cloud/tap/rwx rwx skill install claude -p "Migrate to RWX"codex exec "Migrate to RWX"cursor-agent -p "Migrate to RWX"gemini -p "Migrate to RWX"

For larger teams:

We make it exceptionally easy for teams to migrate over to RWX. Our team has deep experience building and scaling developer infrastructure, and we'll work alongside your engineers to make adoption smooth and fast.

Share this post

Enjoyed this post? Please share it on your favorite social network!

Never miss an update

Get the latest releases and news about RWX with our newsletter.

I am human
hCaptcha