AI writes. Your team decides. Every change leaves evidence.
Rixal governs AI-assisted development: a plan before code, adversarial review by a different provider, and retrievable evidence for every change. It runs on Claude, Cursor, Codex and Grok—and sells you none of them.
Code never crosses the boundary. Rixal keeps decisions, identifiers and evidence—not source code.
Nine stages. A person decides two of them.
Select a stage to see what it produces and who controls it.
A person accepts or returns the plan. Whoever writes the change never approves their own work.
Three architecture decisions, not three features.
We do not sell models
Rixal governs Claude, Cursor, Codex and Grok. Since we do not profit from the model you choose, we can select the least expensive one capable of the task.
We never see your code
Rixal sits above the model call, not in the middle. Code stays on your team's machines; we keep decisions, identifiers and evidence. This is how it is built, not just a policy promise.
Runs on the plans you already pay for
Instead of API keys billed per token, Rixal runs through your team's existing subscriptions and selects by available quota, billing cycle and cost.
What it does today.
Every line has a status. An honest table with half its features in development is worth more than a complete list without statuses.
Two ways to start.
Neither is a buy button. Rixal governs how your engineering team works: it requires implementation, not just installation. Today one path is available and the other has a waiting list, and we clearly identify each.
By project, in your organization
We implement Rixal around your operation: repositories, invariants, review rules and the agents your team already pays for. A project with scope, owners and deliverables—not a per-seat license.
As a service
We are preparing a version where your team signs up and starts without an implementation project. It is not ready: customer isolation is still pending, and we will not open it before that is resolved.
- We email you when it opens, and nothing more
- No campaigns or sales calls
- Unsubscribe in one click
Neither option fits? I want to understand how you built it — someone who uses it every day will reply, not a form.
The longer story, when you want it.
The problem we solve, the order in which we are building it, and what changed this week. None of this is required to decide whether to talk.
What we will not do
- Our own coding agent. Rixal governs the agents you already use. The day we sell our own, we lose the ability to choose the best one for your task.
- Autonomous execution without a person involved. Human control is not a temporary limitation we will remove as models improve. It is the product.
- Store your code. This is not a policy we can change later: it is the reason the system is built this way.