Open-source AI coworkers that each get a computer of their own: a browser, files and tools, with every action decided before it happens and recorded after. Bring any AG-UI agent.
Claude Code starter kit for TypeScript repos
This boilerplate sets up a repository so Claude Code can work from shared rules, commands, agents, and skills. It organizes the flow from clarifying a request to design, implementation, checks, and review, with project-specific context stored alongside your code.
Builders who want Claude Code to follow repo-specific rules and reusable workflows from the first session.
You can run structured Claude Code changes in a repo that already knows its rules, checks, and reusable workflows.
What it does
Project-wide Claude rules
`CLAUDE.md` defines what Claude Code can decide, when it should ask, and how the project works.
Reusable command entry points
`.claude/commands/` provides commands for implementation, design, planning, review, diagnosis, and setup.
Specialized agents
`.claude/agents/` holds roles for repository analysis, design, implementation, testing, and review.
Project skills
`.claude/skills/` stores guidance that Claude loads when it matches the current work.
Quality checks in the workflow
The workflow runs applicable tests, type checks, linting, and build checks before completion.
Skill editing flow
`/create-skill`, `/refine-skill`, and `/sync-skills` help keep project guidance organized and in sync.
Multilanguage setup
The starter kit supports English, Japanese, and Simplified Chinese project files and commands.
How to get it
- 1Run
npx create-ai-project my-project cd my-project pnpm install claude
- 2Run these commands from the project root
npx create-ai-project update --dry-run npx create-ai-project update claude
- 3Once Claude Code is running
/project-inject /implement Add rate limiting to the API
README
AI Coding Project Boilerplate: A Starter Kit for Claude Code
Read this in other languages: 日本語 | 简体中文
Set up a TypeScript repository for structured development with Claude Code. create-ai-project adds a project-level CLAUDE.md, ready-to-use commands, specialized agents, and skills so Claude can take a request through design, implementation, and verification using rules stored alongside your code.
Use it to start a new project and keep its Claude Code setup up to date. Instead of assembling prompts and agent definitions yourself, you get a working development environment that lives in the repository, where your team can version, share, and adapt it.
What this starter kit helps with
- Use
CLAUDE.mdto define project-wide rules, what Claude Code can decide, and when it should ask you - Complete a change with
/implement, from clarifying the request to implementation and verification - Keep straightforward changes lightweight while adding design documents and reviews when the change needs them
- Review the completed implementation to make sure it meets the agreed outcome and repository standards, with no unnecessary changes or serious functional, reliability, or security issues
- Run the repository's applicable tests, type checks, linting, and build checks as part of the workflow
- Record project-specific context and turn recurring team knowledge into reusable skills
- Use the same setup in English, Japanese, or Simplified Chinese
What it adds to your repository
| Path | Purpose |
|---|---|
CLAUDE.md | Project-wide rules, including what Claude Code can decide and when it should ask you |
.claude/commands/ | Entry points for implementation, design, planning, review, diagnosis, and project setup |
.claude/agents/ | Specialized roles for repository analysis, design, implementation, testing, and review |
.claude/skills/ | Development guidance that Claude loads when it is relevant to the current work |
docs/guides/ | Setup, command, and skill-editing guides for project users |
The kit also includes /create-skill, /refine-skill, and /sync-skills so you can add project-specific guidance without maintaining the skill structure by hand.
Quick start
Start a new project
npx create-ai-project my-project
cd my-project
pnpm install
claude
Update a project created with this starter kit
Run these commands from the project root:
npx create-ai-project update --dry-run
npx create-ai-project update
claude
The updater refreshes the managed CLAUDE.md, commands, agents, and skills without replacing your source code or existing package settings.
Once Claude Code is running:
/project-inject
/implement Add rate limiting to the API
/project-inject records the repository-specific information Claude needs, including domain constraints, quality standards, directory conventions, and where to find external schemas or API contracts. You can then use /implement for an end-to-end change.
See the Quick Start Guide for the full setup and first-run walkthrough.
How development runs
flowchart LR
A[Request] --> B[Clarify the outcome]
B --> C[Inspect the repository]
C --> D{Design decisions needed?}
D -->|No| E[Implement directly]
D -->|Yes| F[Design and approve the plan]
F --> E
E --> G[Run checks and review]
G --> H[Complete]
Claude Code first confirms what the change should accomplish and inspects the existing implementation. Straightforward changes can proceed directly. Changes that need product or technical decisions get the necessary design and planning documents before implementation. Before approving a design, Claude checks any facts that could affect the chosen approach against the repository or actual behavior. Any additional experiment is limited to what is needed to make that decision. The workflow then runs the applicable repository checks and reports anything it could not verify.
See Use Cases & Commands for when documents are created, how tests are selected, and what each workflow covers.
Common entry points
| Command | Use it for |
|---|---|
/implement | Take a change from requirements through implementation and verification |
/design, /front-design | Design a change before implementation |
/plan, /front-plan | Turn an approved design into an executable plan |
/build, /front-build | Continue from an approved plan |
/review, /front-review | Review completed implementation against the agreed outcome, repository standards, and security requirements |
/diagnose | Investigate a problem and compare solutions backed by the findings, without changing code |
/project-inject | Record project-specific context and quality standards for future Claude Code sessions |
/create-skill, /refine-skill | Add or improve reusable project guidance |
See Use Cases & Commands for examples and the complete command reference.
Adapt it to your project
Use /project-inject for facts, constraints, and quality standards that apply across the repository. This keeps Claude aware of the project's purpose, conventions, and external sources without repeating them in every request.
When your team has guidance that should apply only to particular work, create or refine a skill instead. The included skill-editing workflow helps decide where the information belongs, reviews the change, and keeps skill metadata in sync. See the Skills Editing Guide for examples and validation guidance.
Language and project configuration
Switch the active Claude Code environment with:
pnpm lang:en
pnpm lang:ja
pnpm lang:zh-CN
pnpm lang:status
The workflows discover the package manager and quality commands from the repository. If your generated project uses different commands, update packageManager and the relevant scripts in package.json.
Guides
License
Files in the repo
- .claude
- .github
- .husky
- bin
- docs
- scripts
- src
- templates
- .gitignore
- .npmignore
- biome.json
- CHANGELOG.md
- CLAUDE.en.md
- CLAUDE.ja.md
- CLAUDE.zh-CN.md
- knip.json
- LICENSE
- package.json
- pnpm-lock.yaml
- pnpm-workspace.yaml
- README.ja.md
- README.md
- README.zh-CN.md
- tsconfig.json
- tsconfig.test.json
- vitest.config.mjs
Discussion (0)
Ask about usage, or say what you built with itSign in to join the discussion.
No comments yet. Be the first to say what this is good for.
More templates
Turn Claude Code into a full game dev studio — 49 AI agents, 72 workflow skills, and a complete coordination system mirroring real studio hierarchy.
A self-organizing Obsidian vault that gives AI coding agents persistent memory. Claude Code, Codex CLI, Gemini CLI.
Production-grade Playwright + TypeScript Scaffold for Agentic Testing. Harness for all major AI coding agents baked in.
Omnia Vault - the all-in-one project brain: an Obsidian LLM wiki, Graphify code graphs, a living plan that triages new videos against itself, and a Claude Code ⇄ Codex relay. Everything your project knows, in one clonable vault.

🎬 Tạo video "so sánh kiến thức" ngắn tự động — HyperFrames + AI voice, 1 template nhiều chủ đề.