Novalux Foreman
A traffic controller for a crew of AI agents
Run a whole crew of coding agents at once — without the work colliding, piling up, or falling apart when it's merged. Local-first, and built on the crew you already run.
- ✓Locks shared files so agents stop colliding
- ✓A merge queue that lands clean work
- ✓Broadcasts context so the pieces fit
Why coordinating agents is the hard part
Press play, or step through with the arrows. No jargon — just what it does and why.
One agent, one job
You hand an AI agent a task. It writes the code and finishes. Simple, and it works.
Everything Foreman gives you
Shared-file locks
When two agents need the same router, config, or migration, Foreman grants it to one at a time and queues the rest — killing the #1 source of merge conflicts.
Merge queue
Finished work lands the moment it's clean; everything behind it rebases automatically, so ten overnight PRs don't rot into ten conflicts.
Context broadcast
The database change, the API contract, and the UI screen are shared across the crew so the slices actually fit together on merge.
Ticket verification
Every agent's output is checked against what its ticket actually asked for, before it's allowed near main.
Built on your crew
Foreman sits on the MCP crew system, worktrees, Kanban board, and review skills you already run — nothing new to learn.
Local-first, solo-scale
No cloud SaaS lock-in. It runs on your machine, for one builder orchestrating many agents — the gap no incumbent serves.
From start to done, step by step
Stage the work
Tickets on the Kanban board become the crew's queue of jobs.
Dispatch the crew
Each agent gets a private worktree and its slice, with shared context broadcast.
Guard the seams
Foreman locks shared files and hands them out one agent at a time.
Verify & land
Work is checked against its ticket, then merged through the queue, rebasing the rest.
What's already shipped, and what's next
Foreman is coming.
It's in active development. Leave your email and we'll tell you the day it ships.