Your terminal, on your organization's rules.
Your engineers keep the terminal they already work in. You keep the rules. A coding AI agent started from the command line is held to the models your organization approved while enforcement is on, what it spends is counted against the project it is working on, and everything it does lands on that project's record.
Signed in by browser, never by key
The terminal shows a short code and you confirm it in the browser. What you get by default is read-only. Anything more, running a project or full access, has to be asked for on purpose and agreed on a separate screen, and the more it can do the sooner it expires. Your sign-in is kept in a file only your own user account can read.
Start an AI agent inside your lines
Start a coding AI agent from the command line and pick a model. The list is grouped by who supplies them: the ones your organization runs, where you need no key of your own, and any provider key it lets you bring. What is on that list is decided by the platform, not by the terminal, so nothing on your laptop can widen it. Anything the coding agent must not do alone is put to a human instead of done.
An AI agent of its own
An AI agent started from your terminal runs on your sign-in, so the work is yours and it is assigned to you. Register an AI agent of its own and the record changes: it becomes the actor and you become the human it acted for. What it may reach is then the overlap between what it was approved for and what you are allowed to do today, and stopping it takes effect on its very next request.
How AI agent access worksThe rest of the day, same place
Clock in and out. Queue work to run in the background and check on it later. Install a package, which asks the approved list first and the package manager second, and if the answer is no, offers to ask a human. Point a coding AI agent you already have at your organization's rules, and unhook it again when you are done.
- alos login, alos projects, alos use
- alos clock, alos task
- alos connect, alos disconnect
Read the command reference in the developer docs.
Also in Studio
- StudioAgent runsGive an AI agent a job, then read back every step of it.Read the page
- StudioAgent securityA name, a limit, and somebody who can stop it.Read the page
- StudioAgent identityA person lets an AI agent in, and can stop it again.Read the page
- StudioWhat an AI agent may doIt never reaches further than its owner.Read the page
- StudioAgent sign-insA sign-in expires sooner the more it can do.Read the page
- StudioApproved modelsWhich models may run, who approved each one, and what it left behind.Read the page
- StudioScans and approved packagesApproved packages at the exact version, watched after that.Read the page
- StudioMCP serverConnect any AI agent, and the rules come with it.Read the page
- StudioSDK and APIOne typed client for the whole platform, given to your team.Read the page