Deploy MCP
Enables AI coding assistants to deploy websites directly to Vercel from the IDE, automating framework detection, build validation, and secure environment variable synchronization, with live status updates.
README
<div align="center">
🚀 Deploy MCP
Deploy any website to Vercel — straight from your AI IDE, in seconds.
Free. Open source. Zero server costs. Built for vibe coders.
Quick Start • IDE Setup • Tools • Troubleshooting • Contributing
</div>
✨ What is Deploy MCP?
Deploy MCP is a free, open-source MCP (Model Context Protocol) server that lets your AI coding assistant (Cursor, VS Code, Antigravity, etc.) deploy your website to Vercel automatically — with full environment variable syncing, build validation, and git safety checks.
Just say "Deploy this" in your AI chat. That's it.
You: "Deploy this website."
AI: ✓ Next.js detected
✓ Build check passed
✓ Environment variables verified (secrets never exposed)
✓ Deployed to Vercel
🎉 Live at: https://my-portfolio.vercel.app
⚡ Quick Start — Deploy in 3 Steps
Step 1 — Run Setup (one-time only)
Open your terminal and run:
npx @tusharjain-19/deploy-mcp setup
This will:
- Ask you for your Vercel Personal Access Token (see how to get it below)
- Save it securely on your machine
- Print the exact config to paste into your IDE
How to get your Vercel Token:
- Go to vercel.com/account/tokens
- Click "Create Token"
- Name it anything (e.g.
deploy-mcp)- Set scope to Full Access
- Copy the token and paste it when the setup wizard asks
Step 2 — Add to Your IDE
Copy the right config for your IDE and add it to your settings file:
Cursor
File location: %APPDATA%\Cursor\User\settings\cursor_settings.json (Windows) or ~/.cursor/rules/cursor_settings.json (Mac/Linux)
{
"mcpServers": {
"deploy": {
"command": "npx",
"args": ["-y", "@tusharjain-19/deploy-mcp"],
"disabled": false
}
}
}
VS Code + Claude Extension
File location: settings.json (Ctrl+Shift+P → "Open Settings JSON")
{
"claude.mcp.servers": [
{
"name": "deploy",
"command": "npx",
"args": ["-y", "@tusharjain-19/deploy-mcp"]
}
]
}
Antigravity (VS Code + Antigravity AI)
File location: settings.json (Ctrl+Shift+P → "Open Settings JSON")
{
"claude.mcp.servers": [
{
"name": "deploy",
"command": "npx",
"args": ["-y", "@tusharjain-19/deploy-mcp"]
}
]
}
✅ Restart your IDE after saving the config.
Step 3 — Deploy!
Open your project folder in your IDE and tell the AI:
"Deploy this website"
The AI handles everything from there:
1. Scans your project → Detects Next.js / React / Vue / etc.
2. Runs build check → Makes sure your code compiles
3. Checks git status → Warns if files aren't committed
4. Scans environment variables → Keys only, values NEVER exposed to AI
5. Syncs missing env vars → Uploads secrets directly to Vercel
6. Deploys to Vercel → Triggers the actual deployment
7. Polls until live → Waits and confirms it's up
8. Returns your live URL 🎉
🔄 Version Control (Git) — Built In
Deploy MCP has full Git support built in. Your AI can:
| What it does | Command to ask AI |
|---|---|
| Check uncommitted changes | "Check my git status" |
| Commit & push before deploy | "Commit my changes and deploy" |
Detect leaked .env files |
"Check if my .env is in git" |
Best practice: Always commit your code before deploying. Ask the AI: "Commit my changes with message 'fix: update homepage' and then deploy"
The git tools work automatically during deployment — if you have uncommitted changes, the AI will warn you and ask if you want to commit first.
🧰 Available Tools
Your AI has access to 16 tools across 4 modules:
📦 Project Tools
| Tool | What it does |
|---|---|
smart_deploy |
All-in-one deploy — detects, checks, syncs env, deploys, polls, and auto-diagnoses failures |
detect_project |
Detects framework (Next.js, React, Vue, etc.) and build config |
check_project |
Validates the project can build successfully |
project_report |
Full pre-flight report — git, build, env, everything |
delete_project |
Permanently deletes a Vercel project (requires your manual approval via OS popup) |
🔐 Environment Variable Tools
Security: Secret values never reach the AI. Only key names are shown. Values go directly from your machine to Vercel.
| Tool | What it does |
|---|---|
scan_env |
Lists all local .env variables (names only, values hidden) |
compare_env |
Shows what's missing on Vercel vs. your local .env |
sync_env |
Uploads missing secrets directly from your machine to Vercel |
create_env_example |
Generates a safe .env.example file with all values blanked out |
validate_environment_variables |
Checks your local env against .env.example requirements |
check_env_leak |
🚨 Detects if any .env files are accidentally tracked by Git |
🌿 Git Tools
| Tool | What it does |
|---|---|
git_status |
Shows branch, uncommitted files, and change count |
git_commit_and_push |
Commits all changes and pushes to remote |
☁️ Vercel Tools
| Tool | What it does |
|---|---|
deploy_to_vercel |
Triggers deployment |
get_deployment_status |
Polls current deployment status |
get_deployment_logs |
Fetches raw build logs for debugging |
diagnose_build_failure |
Parses logs and gives AI-readable fix suggestions |
🛠 Troubleshooting
❌ "Vercel not authenticated" or token error
npx @tusharjain-19/deploy-mcp setup
Re-run setup and paste a fresh token. Make sure the token has Full Access scope.
❌ "MCP not showing up in my IDE"
- Double-check the JSON config you pasted — look for missing commas or brackets
- Fully restart your IDE (close all windows, reopen)
- Test the server manually:
You should see:npx @tusharjain-19/deploy-mcpDeploy MCP server running on stdio
❌ "Build failed" during deployment
Ask your AI: "Diagnose the build failure and fix it" — the diagnose_build_failure tool will automatically parse the error logs and tell the AI exactly what to fix.
Common causes:
- Missing
npm install→ runnpm installin your project - Wrong Node.js version → add
"engines": { "node": "20" }topackage.json - Missing environment variable → ask AI to sync your env vars
❌ "Environment variables not syncing"
- Make sure your Vercel token has Full Access permissions
- Check that the project name you tell the AI exactly matches your Vercel project name
- Run: "Compare my local env vars with Vercel" — the AI will show exactly what's missing
❌ ".env file accidentally in Git"
Ask your AI: "Check if my .env is tracked by git"
The check_env_leak tool will detect it. If found, fix it:
# Remove .env from git tracking (keeps the file locally)
echo ".env" >> .gitignore
echo ".env.local" >> .gitignore
git rm --cached .env
git rm --cached .env.local
git commit -m "fix: remove .env from git tracking"
❌ "Module not found" error locally
npm install
npm run build
❌ "Cannot find Vercel CLI"
Deploy MCP uses the Vercel REST API directly — no Vercel CLI needed. But if you want the CLI anyway:
npm install -g vercel
⚙️ How It Works
┌─────────────────────────────────────────────────┐
│ Your IDE (Cursor / VS Code / Antigravity) │
│ │
│ You say: "Deploy this" │
└───────────────────┬─────────────────────────────┘
│ MCP Protocol (stdio)
▼
┌─────────────────────────────────────────────────┐
│ Deploy MCP Server (runs on YOUR machine) │
│ │
│ ├── detect_project() → Read package.json │
│ ├── check_project() → Run build locally │
│ ├── scan_env() → Read .env names only │
│ ├── sync_env() → Push secrets directly │
│ └── deploy_to_vercel() → Trigger deployment │
└───────────────────┬─────────────────────────────┘
│ Vercel REST API
▼
┌─────────────────────────────────────────────────┐
│ Vercel │
│ ├── Finds or creates your project │
│ ├── Receives encrypted env vars │
│ ├── Builds your site │
│ └── Returns your live URL 🎉 │
└─────────────────────────────────────────────────┘
Security Model
❌ INSECURE (secret touches the AI):
.env → AI Model → MCP → Vercel
✅ DEPLOY MCP (secret never leaves your machine):
.env → MCP Process → Vercel API
↑
AI sees key NAMES only, never values
🔧 Development Setup
Want to run Deploy MCP locally from source?
# 1. Clone the repo
git clone https://github.com/Tusharjain-19/deploy-mcp.git
cd deploy-mcp/mcp-server
# 2. Install dependencies
npm install
# 3. Build TypeScript
npm run build
# 4. Run setup wizard
npm run setup
# 5. Start the server (IDEs do this automatically)
npm run dev
Test All Tools Without an IDE
npx @modelcontextprotocol/inspector npx tsx src/index.ts
This opens a browser UI where you can call every tool manually and see outputs.
Available Scripts
| Script | Command | Description |
|---|---|---|
| Build | npm run build |
Compile TypeScript to dist/ |
| Dev | npm run dev |
Run server directly with tsx |
| Setup | npm run setup |
Run the setup wizard |
📁 Project Structure
mcp-server/
├── src/
│ ├── index.ts ← MCP server entry (all 16 tools registered)
│ │
│ ├── tools/
│ │ ├── smart-deploy.ts ← All-in-one deploy with auto-diagnose
│ │ ├── check-project.ts ← Build validation
│ │ ├── deploy.ts ← Vercel deployment trigger
│ │ ├── diagnose.ts ← Build failure diagnosis
│ │ ├── env-vars.ts ← .env scanning, diffing, syncing
│ │ ├── git.ts ← Git status, commit, leak detection
│ │ ├── logs.ts ← Deployment log fetching
│ │ ├── delete-project.ts ← Safe project deletion with OS prompt
│ │ └── project-report.ts ← Master pre-flight report
│ │
│ ├── vercel/
│ │ └── client.ts ← Vercel REST API wrapper
│ │
│ ├── utils/
│ │ ├── config.ts ← Token storage/retrieval
│ │ ├── filesystem.ts ← File reading utilities
│ │ └── framework-detector.ts ← Framework & package manager detection
│ │
│ └── cli/
│ ├── index.ts ← CLI entry point
│ └── setup.ts ← Interactive setup wizard
│
├── dist/ ← Compiled JS (auto-generated, git-ignored)
├── package.json
├── tsconfig.json
└── README.md
🗺 Roadmap
| Version | Status | Features |
|---|---|---|
| v1.0 | ✅ Done | Project detection, build check, Vercel deployment, status polling |
| v1.5 | ✅ Done | Env scanning (no leakage), env diff, env sync, git tools, project report |
| v1.6 | ✅ Done | Smart deploy (auto-diagnose failures), safe project deletion |
| v2.0 | 🔄 Planned | AI error auto-fix loop, preview deployments, rollback |
| v3.0 | 💭 Future | Netlify/Railway support, GitHub integration, team deployments |
🔄 Automatic Updates
Because your IDE config uses npx -y @tusharjain-19/deploy-mcp, Node.js will automatically pull the latest version from npm every time your IDE starts the server. No manual update commands needed — you always get the latest features and fixes automatically.
🤝 Contributing
Contributions are welcome! This is a free, open-source project built for the community.
# Fork the repo, then:
git clone https://github.com/YOUR-USERNAME/deploy-mcp.git
cd deploy-mcp/mcp-server
npm install
# Create a feature branch
git checkout -b feature/my-feature
# Make changes, then build and test
npm run build
# Commit and push
git commit -m "feat: add my feature"
git push origin feature/my-feature
# Open a Pull Request on GitHub
Please open an issue first for large changes. Bug fixes and docs improvements are always welcome!
🔐 Security
| Principle | How it's implemented |
|---|---|
| Secrets never touch the AI | scan_env returns only key names + classifications |
| No external servers | MCP runs entirely on your machine |
| No overwrite without consent | sync_env skips existing Vercel vars by default |
| Leak detection | check_env_leak scans git ls-files before deploy |
| Token stored locally | ~/.deploy-mcp/config.json — your machine only |
| No hardcoded credentials | Zero hardcoded tokens in the codebase |
👨💻 About
Built by Tushar Jain — tools that bridge the gap between AI and shipping real products.
📄 License
MIT License — 100% free forever. Use it in commercial projects, modify it, distribute it. See LICENSE for full text.
<div align="center">
Built with ❤️ for vibe coders everywhere
⭐ Star this repo if it saved you time!
</div>
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。
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。
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。
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。
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。
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。
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。
mcp-server-qdrant
这个仓库展示了如何为向量搜索引擎 Qdrant 创建一个 MCP (Managed Control Plane) 服务器的示例。