SHELL & RUST · LOCAL AI AGENTS
locals — Concurrent Local AI Coding Agent Orchestrator for Apple Silicon
locals is a macOS command-line orchestrator for running multiple local AI coding agents in parallel across isolated Git worktrees. It manages Apple Silicon unified memory pressure, shared KV prompt caches, and automated branch merging so local agent swarms complete tasks up to 3× faster.
$ brew install bhubbard/tap/locals
Copy
GitHub ↗
Throughput Gain
3.1× Faster
Worktree Clone
0.4 ms (APFS CoW)
Shared KV RAM Saved
62% Reduction
Max Parallel Agents
16 Workers
01. Interactive Workbench
02. Reference Implementation
03. Architecture & Specs
Interactive Apple Silicon Unified Memory & APFS Worktree Scheduler
Concurrent Agent Tasks & Target Branches
Unified Memory & Filesystem Isolation Mode
APFS Copy-on-Write (`cp -c`) + Shared Prefix KV Cache (3.1× speedup)
Git Worktree Isolation + Per-Agent Metal Command Queue
Read-Only OverlayFS + Deterministic Diff Merger
Benchmark Mode: Sequential vs 8-Way Parallel Wall-Clock
Execute Pipeline ▶
Reset
Deterministic Execution Output & Telemetry
STATUS: READY // 0 ERRORS
Shell / System Orchestrator
# Instant zero-disk-cost APFS copy-on-write workspace clone on macOS
cp -cR ./repo ./workspaces/agent-01
# Pin agent inference to shared KV prefix in Apple Silicon Unified Memory
locals run --agents 4 --shared-kv-prefix --isolation apfs-cow --merge-strategy rebase
Production Engineering Specifications
Repository: https://github.com/bhubbard/locals
Live Studio: https://code.brandonhubbard.com/locals/
Portal: https://code.brandonhubbard.com/
Install: git clone https://github.com/bhubbard/locals
Design System: Brandon Hubbard Swiss Modernist (#0A0A0C / 1px Exposed Structural Grid)