How it works
From a request to a change you approve
Your AI team works like a small engineering company: one agent plans, others build, another reviews, and you give the final yes.
Four steps
01
You describe a mission
Say what you want, in your own words. The lead agent breaks it into small tasks and writes down what each one must prove.
02
Builders build
Each task goes to a builder agent, which writes the code, runs the tests and hands in exactly what it made.
03
Reviewers check
A different agent checks the work against what it had to prove. Work that falls short goes back, with the reason.
04
You approve
You see what is ready, try it and approve it. Only what you approve ships.
The team
Every team has the same roles. You choose how many agents it has and which models they use.
- LeadPlans each mission, hands out the tasks and keeps the team unblocked.
- BuildersEach builds one task within its limits and proves that it works.
- ReviewersCheck every task and pass or fail it, with the evidence.
- YouYou make the decisions, approve the work and choose when it ships.
What you get
- Every change reviewed
- Tests run and a second agent reviews each change before you see it.
- You decide what ships
- The team never ships on its own. Each change waits for your approval.
- A machine of your own
- Your agents run on a dedicated machine that we set up and run. Your code stays there.
- Progress while you sleep
- The team keeps working through the night. In the morning you find what is done and what needs you.
Built on orx
Under the hood, orxtrate runs orx, the open-source harness for teams of coding agents. It is the same tool we use to build orxtrate itself.
- Open source at the core. The software that runs your team can be read line by line.
- Review is built in. No change reaches you without a review by a second agent.
- We use it every day. This site and the service behind it are built by orx teams.
Frequently asked questions
What teams ask us before they sign up.
What exactly is orxtrate?
A managed service that gives you a team of AI coding agents. A lead agent turns your requests into small tasks, builder agents implement and test them, and reviewer agents check each one. You see everything in a control panel and approve what ships.
Where does my code run?
On a dedicated machine that we create for you when you subscribe and keep running for you: updates, backups and monitoring are on us. No other customer’s code or agents run on it.
Who can see my code and my work?
You and the people you invite. Your code is never used to train AI models. If our operators ever need to open your account for support, it is recorded in your audit log.
Which AI models do the agents use?
Coding agents such as Claude Code and Codex. Their tokens go apart from the plan, with our credits (the API price plus 20%). Soon also with your own API key (nothing on top from us) or your own Claude or ChatGPT subscription, which stays on your machine alone.
Do I need to install anything?
No. Everything runs on your dedicated machine, and you use it from the browser.
How do I cancel, and what happens to my data?
Cancel at any time; your service runs until the end of the period you paid for. We then export your data, delete it from the machine and shut the machine down. You can export your data whenever you like before that.
How much does it cost?
Solo is 359 USD a month, Team 1,449 and Company 2,299, each with a dedicated machine. The AI models’ tokens are apart.
When can I start?
We are opening to a few customers at a time. Leave your email and we will get in touch as soon as there is room.