Your project gains an AI team.Your company keeps control of it.

An AI team that works inside your project — reads the code, edits, runs commands, tests, queries the database and drives the browser — under your company’s control: scope declared by you, permission by impact level, and every call audited with model, cost and duration.

Yes. Four production systems were built with Kodzonia for third parties: a payments platform and three systems inside an enterprise ecosystem with decades of code. The fifth is the product itself.

For whoever builds software — from scratch or on top of what already exists. The independent dev opening a clean project today who wants to deliver tomorrow; the software house minding code for several clients; the product team with legacy few people understand; the company with hundreds of developers that wants AI with governance — including decided by someone who does not write code: scope, permission and cost are declared and audited, not trusted. Born with Kodzonia or bringing thirty years of code, the AI team is the same — and it grows with yours, not against it.

Yes — and that is what makes the rest possible. You sign in with a Zonialabs account and the project becomes a resource of it: conversations, memories, decisions and audit are available to whoever you authorize, on any machine. Without an account, each person would be an island with an API key in their pocket. Your source code still stays only on your machine.

Two roles. Whoever governs decides the providers, which models are available to the team, at what capability tier each one runs and which ones automatic routing may pick from, the permissions and the limits — tracks per-person cost on the panel, invites by email and revokes access on the spot. A member opens the project and works, with the same history and memories the team built. Governance here is not bureaucracy: it is the power to decide, see and cut — with everything on record.

Whoever manages the team, in one place. The credential lives on your account's server and is injected into the call by the server — the desktop never receives it. Rotating is a single act that holds for everyone; removing someone from the team removes access at the same instant. Nobody needs to paste a key into a config file to get to work.

Yes. Every call becomes a line with model, tokens, cost and duration, attributed to the project and to whoever triggered it. Spend aggregates per member, per day and per project. Team cost stops being an estimated split.

Yes. You connect the provider account your company already uses and consumption is billed there, by the provider. The difference is where the key lives: in your Zonialabs account’s vault, used by the server on every call and never handed to the desktop. Work with another provider? The answer fits in one conversation.

Today: 18 models from Anthropic, OpenAI and Kimi, side by side in the same catalog, each with a verified fact sheet — limits, formats, capabilities; what isn't known stays blank instead of becoming a guess. Switching models rewrites nothing: the tools, the memory, the specialists and the audit belong to Kodzonia, not to the model. And the provider policy is yours: the project only uses the providers you authorize — no others. The catalog is growing: support for DeepSeek, Muse Spark 1.3 (Max) and GLM is on the way, to the same standard. Use another provider? The answer fits in one conversation.

You swap the engine and the project continues exactly where it was. Memory, specialists, workflows and audit belong to Kodzonia, not to the model — a provider outage or a policy change does not stop your team.

Yes, by construction. Conversations, memories, decisions and audit belong to your project. Your code was never anywhere but where you work.

Every model call becomes an audit line with model, tokens (cache included), duration and error, plus the tree of which agent called which. The panel aggregates by day, model and tool, and exports to CSV.

It opens the real project and works in it: reads the code, edits, runs the test suite, reads the error, fixes it and runs again. Before it reaches you, another specialist reviews what was done — code review is a role on the agent team, not a favour someone does at the end of the day. You get the work already reviewed, diff in hand: approve it, ask for changes, or review it yourself whenever that makes sense. And what was learned along the way can become project memory: the rule explained once never needs explaining again, not to it and not to whoever joins later.

macOS today. It is a native desktop app — you install it, open it and sign in with your Zonialabs account; the project shows up with the team’s history. The Windows version is in development: join the list telling us which system you use, and we will let you know when it opens.

It depends on the model and the size of the tasks — which is why the product measures instead of estimating. The panel shows consumption per call, per day and per project from week one, so you find your number with your code, not someone else's average. And FMR — our model-routing technology — uses a light model for what is simple, saving the expensive one for when the task is about to change things.

That is exactly what the beta is. You get in, use it on real code and decide from your own experience. Whoever joins now also talks straight to the people building it and shapes what comes first.

By email, at contato@zonialabs.com, and on the other end is whoever wrote the code — no support tier in between. We answer within one business day.

What belongs to the project — conversations, memories, decisions, audit — lives in your Zonialabs account, under your own access control, and the source code never leaves your machine. If you want to leave, we hand over the project's data. If your legal team wants that in writing, it goes into the contract.

We do. Bring your legal team’s template or use ours — we settle it in conversation, before any access.

The product is used daily on real code, including to build itself, but it is not in open production yet. Whoever joins now follows it closely and shapes the priorities.