tag

TAG

The scaffolding tool that keeps working after day one.

Most scaffolding tools create a project and disappear. TAG stays — generating code into your project as it grows. One binary, Jinja2 templates, built for AI coding agents.

Quick Start

Install

# macOS/Linux
curl -sSfL https://raw.githubusercontent.com/kaikenlabs/tag/main/install.sh | sh

# With Go
go install github.com/kaikenlabs/tag@latest

Scaffold a Project

# From a GitHub template
tag scaffold gh:user/go-api my-service

# From a local template
tag scaffold ./my-template

Generate Code Into It Later

cd my-service
tag generate handler UserAuth
# → creates internal/handlers/user_auth_handler.go

Why TAG

Scaffold and generate in one tool. Most scaffolding tools (Cookiecutter, Yeoman) create a project and walk away. Code generators (Plop, Hygen) only work inside existing projects. TAG does both — scaffold a complete project from a template, then keep generating code into it as your project grows.

Built for AI coding agents. TAG templates include .skills/ files that AI coding assistants understand. Your AI agent can scaffold projects, create generators, and run code generation without manual copy-paste.

Single binary, familiar syntax. One curl to install. No Python, no Node, no runtime. Templates use Jinja2 syntax you already know.

Migrate from Cookiecutter. TAG auto-detects Cookiecutter templates and converts them. Your existing templates keep working.

How It Works

Scaffolding

1. Pick a template — local, GitHub, GitLab, Bitbucket, or any Git repo
2. Answer interactive prompts (or pass variables on the command line)
3. TAG renders every file, runs hooks, and outputs your project

Code Generation

1. Run tag template init in your project
2. Create generators in .tag/ with frontmatter — create files, append to files, or inject at markers
3. Run tag generate <generator> <name> to add code

Features

Commands

Documentation

For AI Coding Agents

If you're an AI agent working with a TAG template, start with .skills/SKILL.md — it has a decision tree, generator anatomy, CLI quick reference, and common pitfalls in an LLM-optimized format. Detailed reference and recipes are in .skills/reference.md and .skills/recipes.md.

License

MIT