CAS runs a supervisor that plans your work, then spawns workers to build it. We call this a factory.
Your request becomes a structured plan — tasks, dependencies, and priorities. Nothing starts until the plan makes sense.
Each agent claims a task and works in its own git worktree. No stepping on each other's code. No waiting.
Workers open pull requests when they're done. You review the code, merge what's good, and give direction on the rest.
The systems that make multiple agents actually useful on a real codebase.
Agents remember past sessions — bugs fixed, patterns learned, your preferences. New sessions start smart, not blank.
Work breaks into tasks with priorities and dependencies. Agents claim what's ready and skip what's blocked.
When an agent makes a mistake and you correct it, CAS creates a rule so it never happens again.
Teach an agent how to deploy, test, or lint once. Every agent on your project gets the skill.
Each agent works in its own copy of the repo. No file locks, no conflicts, no waiting on each other.
What one developer's agents learn, the whole team's agents know. Push context to the cloud, pull it anywhere.
Runs entirely on your machine. Your code never leaves your laptop. Cloud sync is optional.
curl -fsSL https://cas.dev/install.sh | sh