Per-user sizing
Size environments per user for the right balance of speed and cost.
Keeping a worktree warm makes the next workspace on that repo start faster, but every warm worktree also costs disk space or cloud resources. Proliferate lets each person tune that trade-off for themselves instead of picking one setting for a whole team.
Ideal worktrees per repo
In Settings → Pruning, set the number of worktrees you want to keep materialized per repo, from 10 to 100 (the default is 20). This is a personal preference: everyone on a repo can set their own number based on how they work and how much local disk or cloud headroom they're comfortable using.
Settings → Pruning, ideal worktrees stepper
The 'Ideal worktrees' stepper control (minus/count/plus, range 10-100) with helper text describing composer pressure and cleanup behavior.
Go above your ideal count and two things happen: the pressure indicator in the chat composer starts to warn you, and Proliferate's automatic cleanup becomes more willing to prune old, clean, unused worktrees for that repo back down toward the target. If you're signed in, this same number is pushed to your cloud sandbox's retention policy, so local and cloud cleanup follow the same target.
Reading the pressure indicator
A small ring sits in the chat composer and reflects the runtime behind your current workspace:
- Local: how many materialized worktrees exist for the repo, compared to your ideal count.
- Cloud: CPU and RAM usage in your sandbox, compared to an ideal ceiling Proliferate maintains for it.
The ring's color tells you how close you are: green while you're comfortably under, amber once you pass about 80% of the limit, and red once you're at or over it. Click it any time to open the full worktree details for that runtime.
Composer runtime pressure indicator
The pressure ring in the chat composer, showing its tooltip with worktree count vs. ideal (local) or CPU/RAM percentages (cloud), plus the color states for comfortable, warning, and over-limit.
Why per-user, not per-org
Sizing lives with the person, not the repo or the organization, because the right number depends on how someone actually works: someone juggling many small parallel tasks wants more warm worktrees than someone doing one focused stream of work at a time. There's no single number that's right for everyone on a team, so Proliferate doesn't force one.
If you're not sure where to start, leave the default. Raise it if you frequently see cleanup remove a worktree you wanted to keep around; lower it if you'd rather trade a little startup latency for less disk or cloud usage.
Where this fits
This setting only controls how aggressively idle worktrees get cleaned up. It never overrides the safety checks described in Lifecycle & storage: a dirty checkout, a live session, or anything still running is left alone regardless of how far over your ideal count you are.