spec-kitty/spec-kitty

Spec-Driven Development with organizational governance. Specs tell AI agents what to build; Charter governs how they build it. Git-native missions, enforceable workflows, and consistent engineering standards across AI coding agents and harnesses.

What it solves

AI coding sessions often lose requirements, decisions, and acceptance criteria, making it hard to run multiple agents in parallel or move from casual vibe coding to a repeatable process. Spec Kitty gives teams a way to turn product intent into a governed, repo-native workflow with clear work-package boundaries, visible progress, and review gates, so the repository remains the single source of truth.

How it works

Spec Kitty is a CLI that follows a defined pipeline: spec -> plan -> tasks -> next -> review -> accept -> merge. You start by initializing a project for a specific agent (like Claude Code, Codex, Cursor, Gemini, Copilot, Windsurf, OpenCode, and others), then use slash commands or skills to create specs, plans, and work packages. The tool stores all mission artifacts under kitty-specs/, tracks work packages through lifecycle lanes (planned, in_progress, for_review, approved, done), and runs agents in isolated git worktrees under .worktrees/. A dispatch command loads governance context and opens an operation record, and the runtime picks the next action until the mission is ready for review and merge. Completed missions generate a retrospective by default.

Who it’s for

Teams building software factories who need repeatable inputs, clear work boundaries, isolated execution, and review gates around AI coding agents. It suits people coordinating multiple agents or developers on large codebases, or those moving from ad hoc vibe coding to a disciplined spec-driven workflow. It is overkill for one-off edits or tiny scripts, and it requires Git.

Highlights

  • Repo-native governance: specs, plans, work packages, acceptance criteria, review state, and merge decisions live in Git.
  • Parallel agent execution via isolated git worktrees, avoiding branch chaos.
  • Supports many AI coding agents (Claude Code, Codex, Cursor, Gemini, Copilot, Windsurf, OpenCode, etc.) via slash commands or skills.
  • Explicit workflow loop: next -> review -> accept -> merge, with an audit trail.
  • Local-first and open source (MIT), with optional hosted sync and tracker integrations that are not required.
  • Every completed mission automatically produces a retrospective, enabling learning across missions.

Related

  • Project
    github/spec-kit

    Spec Kit is an open-source toolkit that provides structured, repeatable processes for AI coding agents to handle feature development, bug fixing, and project assessment.

  • Project
    kaochenlong/spectra-app

    A Spec-Driven Development tool consisting of a desktop app, CLI, and AI agent skills to visually track and manage specifications, change proposals, and tasks.

  • Project
    clawplays/ospec

    A spec-driven, agentic workflow framework for AI coding agents that persists requirements, plans, and verification evidence directly in the project repository.

  • Project
    shuxueshuxue/Spexcode

    A Git-backed spec tree and session manager for coding agents that prevents documentation drift by anchoring specs to code and blocking changes that break the spec-code link.

  • Project
    codeaholicguy/ai-devkit

    AI DevKit is a Node‑based, Apache‑2.0‑licensed tool that adds a shared configuration, console UI, SQLite‑backed memory, cross‑agent messaging, and composable engineering “skills” (requirements, design, TDD, verification, security review, etc.) to multiple AI coding assistants (Claude Code, Gemini‑CLI, Codex‑CLI, Cursor, Copilot, etc.). It lets developers run, supervise, and coordinate these agents from a single control plane, reducing scattered setups and enforcing disciplined development workflows.