Show HN: BranchBox – isolated dev environments per feature (github.com)

BranchBox is an open-source engine that provisions fully isolated, per-feature development environments so engineers — and AI coding agents — can run multiple features in parallel without collisions. Each feature gets a dedicated Git worktree, devcontainer, Docker Compose project and network, database, ports, .env, and optional Cloudflare/SSH tunnels; shared credentials can be mounted explicitly. It’s a full-stack environment (Rails/Node/Python, containers, DBs, VS Code/Cursor devcontainers) rather than a lightweight sandbox, and is available via Homebrew or a direct install script. For the AI/ML community, BranchBox addresses a growing pain point as agentic workflows proliferate: LLMs performing migrations, refactors, or experiments could otherwise leak state, collide on ports, or drift devcontainers. An always-on BranchBox Agent provides observability (feature events, heartbeats, stack metadata, control-plane connectivity) and safe operations (start/tear worktrees, run minimal spikes, apply prompt seeds, reuse credentials). The technical implication is predictable, reproducible local compute per feature — enabling simultaneous human and agent development, reducing “I broke main” incidents, and making parallel experimentation and CI-adjacent agent workflows practical and low-risk.