onboard-mcp
Checks your local development environment for common issues like wrong Node version, missing Docker, missing git config, missing .env keys, and busy ports, and provides fixes.
README
onboard-mcp
Your dev environment, fixed by asking your AI assistant.
An MCP (Model Context Protocol) server that checks the things that silently break local development: wrong Node version, Docker not running, missing git config, missing .env keys, busy ports. Ask your AI assistant "is my environment ready?" and get a pass/fail report with the exact fix for everything that's broken.
Why
Every developer has lost an afternoon to "it works on my machine": the app fails because Node is one major version off, or the port is taken by a zombie process, or .env.local is missing a key that was added last sprint. On my team, environment setup issues cost every new developer days.
This server turns that debugging into one question in your AI chat:
You: run the doctor on my project, it needs ports 3000 and 5432
Assistant:
[PASS] node-version: Node v20.11.0 satisfies the project requirement (20). [FAIL] docker: Docker is installed but the daemon is not running. Fix: Start Docker Desktop, or on Linux run "sudo systemctl start docker". [PASS] git-config: Git is installed and user.name / user.email are configured. [FAIL] env-files: .env.local is missing 1 key(s) from .env.example: REDIS_URL. Fix: Add the missing keys to .env.local. Ask a teammate or your secrets manager for the values. [PASS] ports: All required ports are free: 3000, 5432. 2 failing, 0 warning(s). Fixes listed above.
Tools
| Tool | What it checks |
|---|---|
doctor |
Runs everything below and returns one report |
check_node_version |
Current Node vs .nvmrc or package.json engines |
check_docker |
Docker CLI installed and daemon running |
check_git_config |
git installed, user.name / user.email set |
check_env_files |
.env / .env.local vs .env.example (compares keys only, never reads your values into the response) |
check_ports |
Whether the TCP ports your project needs are free |
Requirements
- Node.js 18 or newer — required for every client, since the server runs via
npx onboard-mcp. Check withnode --version; install from nodejs.org if needed.
That's the only universal dependency. npx fetches the published package automatically — no clone or build required. The Claude Code steps below additionally need the claude CLI (npm install -g @anthropic-ai/claude-code); other clients do not.
Quick start
Claude Code
Requires the Claude Code CLI. Install it first if you don't have it:
npm install -g @anthropic-ai/claude-code
Then register the server:
claude mcp add --scope user onboard -- npx -y onboard-mcp
(--scope user makes it available in every project; omit it to register for the current project only.) Verify with claude mcp get onboard.
Or clone and build locally:
git clone https://github.com/developerpankajdixit/onboard-mcp.git
cd onboard-mcp && npm install && npm run build
claude mcp add onboard -- node /absolute/path/to/onboard-mcp/dist/index.js
VS Code (GitHub Copilot agent mode)
Add to .vscode/mcp.json in your project:
{
"servers": {
"onboard": {
"command": "npx",
"args": ["-y", "onboard-mcp"]
}
}
}
Cursor
Settings → MCP → Add new global MCP server, or add to .cursor/mcp.json:
{
"mcpServers": {
"onboard": {
"command": "npx",
"args": ["-y", "onboard-mcp"]
}
}
}
Claude Desktop
Add to claude_desktop_config.json:
{
"mcpServers": {
"onboard": {
"command": "npx",
"args": ["-y", "onboard-mcp"]
}
}
}
Then ask: "Run the doctor on /path/to/my/project, it needs ports 3000 and 5432."
Any other MCP-compatible client works the same way: point it at npx -y onboard-mcp over stdio.
Design notes
- Fixes, not just failures. Every failing check returns the exact command to fix it, so the assistant can offer to run it.
- Secrets never leave your machine. The env check compares key names only. Values are never included in any tool response.
- Pure check functions. Every check is a plain async function with injectable dependencies, tested without mocking the MCP layer (23 unit tests, CI on Linux and macOS across Node 18/20/22).
Development
npm install
npm test # vitest
npm run build # tsc
npm run dev # run the server from source
Roadmap
fix_*counterpart tools that apply the suggested fixes after confirmation- Configurable check list via an
onboard.config.jsonin the project root - Database connectivity checks (Postgres, Redis) behind a flag
License
MIT. Built by Pankaj Dixit, based on an internal MCP onboarding tool that cut new-developer setup from 2 weeks to under 30 minutes.
推荐服务器
Baidu Map
百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。
Playwright MCP Server
一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。
Magic Component Platform (MCP)
一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。
Audiense Insights MCP Server
通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。
VeyraX
一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。
graphlit-mcp-server
模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。
Kagi MCP Server
一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。
e2b-mcp-server
使用 MCP 通过 e2b 运行代码。
Neon MCP Server
用于与 Neon 管理 API 和数据库交互的 MCP 服务器
Exa MCP Server
模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。