How Constraint works.
What runs where, what your team shares, and where your people step in.
A desktop app on each computer.
Constraint is a desktop app your whole team shares. Each person runs it on their own computer. Sentries, scheduled checks that spot recurring work, run while the app is running.
Work lives on tickets.
Your team and its agents work the same tickets, in the open. The tickets live in your ticketing service.
One agent coordinates.
Talk to one agent, the VP. It files the tickets and puts an agent team on every one, and they work in parallel. A ticket goes to one role, or to an agent team that runs AI roles in sequence. In the app, a team like this is a pod.
Roles, models and skill versions.
You choose your roles, each with its own job, skills and model, and group them into agent teams. Skills come in versions, and each role pins the version it uses.
What's shared, and what stays on each computer.
Your team shares its departments and their AI roles, each with its job, skills and model, plus the sentries that watch for new work. Each person's AI sessions and their records are kept on their own computer. Your AI service and connected tools receive what each session sends them.
Connect your own accounts.
Connect your own AI account and choose from its models. Through our integration provider, teams can connect the services they already work in. An administrator approves each connector, and each person connects their own accounts.
Where your people step in.
The app's Home shows each person the decisions that are theirs to make. A worker asks there and waits for the answer, and your settings decide what needs permission. Export a session's trace, and a teammate imports it into their app and takes over from your decisions.