How it works
Guide. Let it work. Measure. Steer.
x-Arbiter governs the collaboration between your team and its AI, not the model itself. This is the loop it runs on every piece of work.
01 · GUIDE
It gives the AI your project and you.
Before the model writes, x-Arbiter gives it what your project is for — its direction, its standards, its history — and how the person directing it works.
02 · WORK
The model does its work.
x-Arbiter doesn’t restrict the model. It goes where it judges it needs to go, in the tool your team already uses, with whichever model you chose.
03 · MEASURE
Every commit is measured against the project.
x-Arbiter checks whether what came back fits your project. The judgment is tested on your own work before it counts, so you know how far to trust it.
04 · STEER
Drift is named, to the AI and to you.
When the work drifts, x-Arbiter says where and what to move toward, so the next step comes back to what you asked for. You rule on every verdict, and every one is kept in your repository.
Setting it up
It runs alongside the AI you already use — nothing to instrument, no pipeline to configure. Licensed teams install it from our private registry; the public install opens at release.
Install in any project
One command adds x-Arbiter to your existing Claude Code or Cursor setup. It works inside the workflow you already have — there is no new tool to learn.
Not yet installable. The install command is intentionally not shown — it would not work today, and a command that fails is worse than no command. It will appear here at release.
Declare what you are building
An onboarding pass records your vision, design laws, and ethical commitments. This is the part no model can infer for you, and it is what the AI is given before it writes and what its work is measured against.
Work as usual; rule on what it finds
Your team works with its AI as it does today. When x-Arbiter names drift, you decide: your rulings count over the judge’s verdicts, and every one is kept in your own repo.
What’s inside
Specialist reviews
On request, specialist reviews of your code and product — quality, experience, business, ethics and more — with every finding cited to the file and line it is about.
Rules you decide on
When the same mistake keeps coming back, x-Arbiter proposes a rule. It becomes permanent only with your yes, and a finding you dispute never counts toward one.
No cold starts
Each session picks up where the last one stopped: what was built, what was decided, and what comes next.
A dashboard for every project
Measures, findings, sessions and decisions across all your projects, visible to your whole team without installing anything.
Plugs into any codebase
Greenfield or decades-deep legacy — x-Arbiter installs into the project you already have and never overwrites your files. No migration, no rewrite, no adoption cliff.
Your stack, not ours
You choose the model — Claude, OpenAI, Gemini, xAI, or a local one — and x-Arbiter works with the tools your team already uses. It conforms to your stack; it never dictates it.
Bring it to your team.
x-Arbiter is licensed directly to businesses. Tell us about your team and what you’re building, and we’ll show you what it measures on one of your projects.
Not yet publicly released; licensed directly.