AI coding agent operations
The implementation details between “open more agents” and “merge software you can trust”: worktrees, runtime ownership, orchestration, benchmarks, datasets, and tools I use myself.
Research and field guides
4 maintained resourcesHow to Run Multiple Coding Agents in Parallel Without Collisions
A field-tested workflow for running Codex, Claude Code, Gemini CLI, and other coding agents in parallel using Git worktrees, resource isolation, a ready queue, and a short-lived integration lease.
Operating model →GUIDE 002Git Worktrees for AI Coding Agents: A Practical Setup Guide
A copy-pasteable Git worktree workflow for running Codex, Claude Code, Gemini CLI, and other AI coding agents without file collisions or destructive cleanup.
Setup and commands →ARCHITECTURE 003Coding Agent Orchestration: Architecture for Parallel AI Development
A practical architecture for coordinating multiple AI coding agents across task ownership, runtime resources, observability, validation, and integration.
System design →DATASET 004Parallel Coding Agent Terminal Benchmarks from GridBash
Reproducible release-mode microbenchmarks from GridBash showing where terminal orchestration spends CPU under large multi-agent grids, with raw JSON and CSV data.
Performance evidence →Use the system
Free / local / no signupParallel Coding Agent Worktree Planner
Generate isolated Git worktree commands, branch names, ports, runtime namespaces, and handoff checklists for multiple coding agents.
Generate a workflow →OPEN SOURCEGridBash
A cross-platform Rust terminal grid for operating real Codex, Claude, Gemini, and other CLI sessions with worktree isolation and prompt routing.
Run the agents →