Running git work-trees with Portless & Conductor

We let Conductor create a git worktree for each task, then start the dev server inside it through Portless. Portless reads the branch name and registers a subdomain for that worktree, so two branches can run side by side at different local URLs instead of fighting over localhost:3000.

Portless is a local proxy that maps a readable .localhost domain to whatever port your dev server lands on. Each worktree gets a free port in the background, and you only ever open the named URL, so you never have to remember which branch is on which port.

Conductor runs several coding agents at once, each in its own git worktree on its own branch. That keeps their changes isolated from each other and from your main checkout. Portless covers the part Conductor doesn't, which is giving each of those worktrees a dev server you can actually open.

The route only exists while that worktree's dev server is running. Portless adds it when the server starts and drops it when the server stops, so start the dev server from inside the worktree and the URL comes back. It isn't a port clash, since every worktree gets its own subdomain and port.

No. You set Portless up once and point your dev script at it. After that, every new worktree picks up its URL from the branch name when you start the dev server, with no extra config per branch.