PILLAR POST

Ship Code From WhatsApp: How CloudAxis Coding Works

Most AI coding tools live inside your editor and hand you a diff to trust. CloudAxis Coding runs in the cloud, proves the change works, and delivers a pull request you approve from your phone. Now in public beta.

9–12 min read

You are away from your desk when you notice the pricing page still says "$29" but you changed the plan to $39 last week. Normally that is a context-switch: open the laptop, pull the repo, find the file, make the edit, run it locally, push, open a PR. Ten minutes of setup for a one-line change.

With CloudAxis Coding you send one message — "update the pricing page to $39/mo" — and a few minutes later you get a reply with a pull request, a live preview of the changed page, and a screenshot. You tap approve. It merges the exact commit you looked at. You never opened an editor.

This post walks the whole loop end to end, and is honest about what it does and does not do yet — because it is a public beta, not a finished promise.

What CloudAxis Coding is (and is not)

CloudAxis Coding is a cloud coding agent: you describe a change, and it produces a reviewable pull request on your GitHub repository. It is part of the same AI cloud computer that runs your other agents — the coding agent just gets its own isolated machine to work on your code.

It is not an editor plugin and not an autocomplete. There is no IDE to open, no diff to babysit inside a tool you still have to run yourself. The output is a pull request with proof attached, and nothing merges without your explicit approval.

The loop, step by step

Behind the one-line message is a sequence designed so that by the time a human sees anything, the work is already done and checked.

1. A private cloud machine, per task

Each coding session gets its own isolated cloud computer — a sealed sandbox, separate from every other user and from your other agents. Your repository is cloned into it with a short-lived, repo-scoped GitHub token (about an hour, not a standing key). When the task ends, the machine is stopped and reaped. Nothing lingers.

2. The agent reads, edits, and tests

The agent detects your stack, installs dependencies, and works in small steps: read the relevant files, make a surgical edit (replace a snippet, apply a patch — not blind full-file rewrites), then run your tests to check itself. It repeats until the task is done or it hits a bounded step limit. Every step is logged to a live timeline you can watch.

3. It proves the change actually works

This is the part editor copilots skip. For a web app, the agent starts your dev server inside the machine, opens the page in a real headless browser, and captures screenshots across desktop and mobile viewports plus any console errors. It attaches the test result too. You do not merge a diff and hope — you see it render and pass. Read more in proof-of-work pull requests.

4. A pull request, delivered to you

The agent pushes the branch and opens a pull request whose body carries the task, a summary, the changed files, and the proof bundle. Then it tells you — in your CloudAxis inbox and, if linked, on WhatsApp — with the PR link, the preview link, and a one-tap way to ship it.

5. You approve; it merges exactly what you saw

Merging is a human decision, always. You reply "YES" (with a short code) or tap approve. The merge is pinned to the exact commit you reviewed — if the branch changed after you looked, the merge is refused rather than shipping something you did not see.

Three ways to start a change

The same engine runs no matter how the request arrives:

What it handles well today — and what is still rolling out

Being a public beta, here is the honest picture. The full "preview + screenshots + tests" proof is strongest for web apps — Node front ends (Next.js, Vite, and friends) and Python web apps like Django are the sweet spot today.

For backend services and libraries that have no dev server to preview, the proof is your test suite rather than screenshots — still real evidence, just not a picture. And a handful of toolchains (Go, Rust, Ruby, PHP) are not yet baked into the build machine, so those repos are on the roadmap, not ready this week.

We would rather tell you that than have you file a change on a Rust service and get a shrug. Start with a web app or a JS/Python repo and it will feel like magic; bring an exotic stack and you are helping us prioritize what to add next.

Who this is for

Founders and small teams who own a codebase but do not want to context-switch into it for every small change. Agencies shipping client sites. Solo builders who want to move a task forward from their phone. If you can describe the change and read a preview, you can ship it — the agent handles the editor part.

It is still your code and your review. CloudAxis Coding is the teammate that does the fiddly part and shows its work; you stay the one who says "ship it."

Frequently asked questions

Does it merge code on its own?

No. Merging always requires an explicit human approval, and the merge is pinned to the exact commit you reviewed. If the branch changes after you approve, the merge is refused.

Which languages work today?

The full preview + screenshot proof is strongest for web apps — Node front ends and Python web apps like Django. Backend/library repos get test-based proof instead of screenshots. Go, Rust, Ruby, and PHP toolchains are on the roadmap.

Where does my code run?

In an isolated, single-use cloud machine cloned from your GitHub repo with a short-lived, repo-scoped token. It is stopped and reaped when the task ends; the durable record is the pull request on your own GitHub.

Is it live?

CloudAxis Coding is in public beta. You can try it on your plan today; expect rough edges and tell us where they are.

Related reading in this series
Proof-of-work pull requests · Approve-to-merge from your phone · CloudAxis Coding (beta)