Tooling · internal system
Parallel Dev Environment
Named URLs instead of remembered ports, and a fully isolated environment per branch so parallel agents stop colliding.
Live
No public endpoint
Dev Env
The problem it solves
Several agents working on the same codebase at once is where the leverage is — and also where two of them bind the same port, share one database, and corrupt each other's state. Remembering which of fifteen services lives on which port number is its own small tax.
How it works
01
Branch
a new worktree per line of work
→
02
Isolate
own ports, database, environment, session
→
03
Name
a stable readable URL, not a port number
→
04
Run
agents work in parallel without contention
→
05
Reap
the environment disappears with the branch
Under the hood
- Each worktree gets its own database container and environment file, so a migration in one branch cannot touch another.
- A local proxy maps readable names to services, which removes an entire category of 'wrong port' mistakes.
- Production is pinned to a dedicated checkout — a build never ships whatever happened to be uncommitted in a shared working tree.
- A board shows every live environment, so nothing is left running and forgotten.
Why it matters
Parallel agents are only faster if they are not fighting. Isolation is what converts concurrency into throughput.