A practical home for working projects

Why this exists

Putting your own code somewhere should be the easy part. It stopped being the easy part.

GitHub belongs to Microsoft now, and it shows. AI is pushed at you whether you asked for it or not, its outages are everybody's outages on the same afternoon, and the roadmap answers to a company whose priorities are not yours. GitLab has the features and none of the polish — getting anything done in it is a job on top of the job.

Codeberg was the independent answer, right up until July 2026, when its members voted 358 to 144 to add this to § 2 (1) 7 of their Terms of Use:

You must not share projects that mostly consist of code written by "generative AI"-tools (including services such as Claude, OpenAI Codex).

Fall on the wrong side of that and your project is removed. Do it again and your account can go with it.

And nobody can tell you which side you are on, because "mostly" is not a line, it is an opinion — formed by somebody else, about work you have already published, after you have already published it. There is no percentage to stay under. No way to check before you upload. Just a vote that happened without you, deciding which tools you were allowed to have open while you worked.

That is a strange thing for a code host to be doing. Nobody is auditing how much of your project came out of Stack Overflow, or a framework generator, or twenty years of muscle memory. A tool became controversial, and a platform decided your work was suspect for having touched it.

So this one does not have that rule

CodeRook will never remove your project because of how it was made. Not now, not at a threshold, not after a vote. There is no percentage, no review of your commit history, and nothing for you to guess at before you press upload.

You are not required to tell anyone whether AI helped, and we will never ask. If you want to say so there is a badge you can switch on, and it is yours to publish. If you would rather not, that is the end of it — nothing here reads silence as an admission, and a single generated line is nobody's business but yours.

AI is a tool. It helps people do their work, the same as every other tool in the chain, and it is where this industry is going whether or not any particular platform approves. A rule against it is not a rule about quality. It is a rule about who is allowed to publish.

This is a position taken deliberately while others went the other way, and it is not going to quietly change once there are enough users to make it awkward. It is written into the terms, not just onto this page.

Who runs it

CodeRook is built and operated by ACCA Gaming Productions, the name CodeRook and VoxAI both sit under. It is independent — no shareholders, no investors, and nobody to answer to about how fast the numbers grow. There is no quarter in which the incentives change.

That independence is not a boast, it is the reason the promises on this site are possible to keep. Your projects are never used as training data and are never sold. If the service itself were ever sold, what changes hands is the service — the code stored on it belongs to the people who put it there and does not transfer with it.

We keep it current and we help the people who use it. Write to support@coderook.com.

Upload what you actually have

A real project is not just source code. It is the texture pack, the sample library, the trained model, the installer, the raw footage, the twelve-gigabyte build nobody else will host. Push any of that at most code hosts and you hit a wall — a file size limit, a rejected extension, or a separate paid add-on to store the parts of your own work that happen to be large.

Here there is no approved-file-types list. You are paying for space, so use the space. The shape of a file has no business deciding whether your project is allowed to exist in one piece.

That is what CBX is for. Keeping large projects is a compression problem before it is anything else, and the formats everyone reaches for were designed for a different decade — so we wrote one, rather than asking you to split your work across three services and remember which part went where.

What you control

You decide who sees a project. Separately, you decide how it may be treated by machines — AI crawlers and training scrapers, search engine bots, mirroring and archiving tools, dependency and security scanners, and anything else automated that arrives without a person behind it.

Four answers, set per project: whether machines may read it, contribute to it, download it, or make requests against it. They are yours to set, they are off by default, and they travel with the project rather than sitting in a policy page nobody reads.

How it is paid for

Subscriptions. Storage and servers cost money, and that money comes from people paying to use the service — not from anything done with the projects stored on it. It is worth being explicit about that, because it is the reason we can promise never to use your work as training data or sell it on: there is no version of the business that depends on doing so.

How it works

Project metadata, accounts, permissions, and immutable version history live in PostgreSQL. Content-addressed project objects live in Cloudflare R2. The desktop application, the command line, and the website all publish through the same authenticated API.