{"author":"johannesjo","children":[],"created_at":"2026-02-22T22:42:40.000Z","created_at_i":1771800160,"id":47115578,"options":[],"parent_id":null,"points":2,"story_id":47115578,"text":"I kept running into a structural problem with AI coding tools.\nYou try one agent. It makes changes. You try another. Now you're manually diffing branches and cleaning up experimental commits.\nThe tools are powerful, but the workflow isn\u2019t designed for comparison.\nSo I built Parallel Code.\nIt runs Claude Code, Codex, and Gemini CLI natively (no API wrapper, no feature abstraction), but isolates each agent in:\nits own terminal\nits own Git worktree\nits own feature branch\nThe workflow becomes:\nSpawn N worktrees\nAssign one AI agent per worktree\nLet them implement independently\nCompare diffs\nMerge the best result\nInstead of sequential trial-and-error, you get parallel exploration with safe isolation.\nWhy not just use tmux?\nYou can \u2014 but you still manually manage:\nbranch creation\nworktree cleanup\nsession naming\ncontext switching\nParallel Code automates the Git structure and keeps the sessions visually organized.\nThis is aimed at people already using AI CLIs heavily.\nI\u2019m curious: Has anyone else moved toward multi-agent parallel workflows instead of single-agent iteration?\nGitHub: https://github.com/johannesjo/parallel-code","title":"Show HN: Parallel Code \u2013 Running multiple AI agents in parallel with worktrees","type":"story","url":null}