spec-kitty

Why we included this project

Running several AI coding agents in parallel tends to scatter requirements and decisions across sessions, and acceptance criteria often get lost in the shuffle. Spec Kitty fixes that by making the repository the source of truth: specs, plans, work packages, review state, and merge decisions all live in Git, and each agent works in an isolated worktree instead of colliding on one branch. It is a governance layer rather than another coding assistant: humans define the spec and acceptance criteria, agents implement inside traceable worktrees, and reviewers accept, reject, or merge with an audit trail. That makes it a good fit for teams moving from vibe coding to a repeatable, reviewable process, especially when several agents like Claude Code, Codex, Cursor, Gemini, or Copilot need clear work-package boundaries. For one-off edits or tiny scripts it is overkill, but for governed multi-agent delivery it gives you a durable operating system where the repository stays the source of truth.