I built a tool with Claude Code to solve the biggest pain point of using Claude Code at scale

There's an irony here: Claude Code helped me build the tool that fixes Claude Code's worst coordination problem.

**The problem:** When you run multiple Claude Code sessions on the same codebase, they don't know about each other. They'll happily edit the same files, introduce conflicting patterns, and leave you with a merge nightmare.

**What I built:** `ruah` — a CLI that sits between your Claude Code sessions and the repo. It ensures:

- Each session gets an isolated worktree (no shared checkout) - File scopes are declared upfront (overlapping claims rejected) - Changes merge back in dependency order (no manual conflict resolution) - Task artifacts are captured (you know exactly what each session changed)

**Why I built it *with* Claude Code:**

Claude Code is absurdly good at coding individual features. But ask it to *coordinate* three copies of itself on the same repo and it has no framework for that. That coordination layer doesn't exist yet in any coding tool — so I built it.

C