Thanks to visit codestin.com
Credit goes to github.com

Skip to content

areebahmeddd/MCP-Box

🧰 MCP Box

MCP Box helps you discover, deploy, and test MCPs in isolated sandboxes. It includes:

  • A friendly CLI to initialize metadata, run security scans, push to a registry (S3), search, and configure popular AI clients (VS Code, Cursor, Windsurf, Claude, ChatGPT)
  • A FastAPI backend to list/get/create MCP servers with optional pricing and security reports
  • An AWS Lambda worker that executes MCP servers on demand directly from their Git repositories

Built with Python (FastAPI, Click), S3 (registry), and optional scanners.

Why this project:

  • There’s no centralized MCP registry to discover all MCPs, and many lack clear usage docs.
  • MCPs on our platform pass a 5‑step security/quality check (SonarQube, Bandit, GitGuardian) to reduce vulnerabilities and promote best practices.
  • Unlike MCPs that run locally on your machine, MCP servers here execute in sandboxed environments and return responses securely.

For setup and deployment, see docs/INSTALL.md.

Key Features

  • Central MCP Registry: S3‑backed registry with per‑server JSON for easy discovery and portability.
  • Sandboxed Execution: MCP servers run in isolated environments and return responses securely.
  • Security Pipeline (5‑step): SonarQube, Bandit, and GitGuardian checks with a unified report.
  • One‑Command Publish: mcpbox push scans, discovers tools, and uploads a unified record to S3.
  • Client Auto‑Config: mcpbox pull --client cursor|vscode|... writes correct MCP config pointing to the Lambda endpoint.
  • Tool Discovery: Regex‑based discovery across Python code and optional Node package.json definitions.
  • HTTP API: FastAPI routes for listing, fetching, and creating MCP records; health check and Swagger docs.
  • Razorpay Payments: Required payment flow to create/verify orders and fetch payment status.
  • Lambda Executor: Fetch repo ZIP, install deps to /tmp, run entrypoint, and stream response.

NOTE: The Lambda executor currently supports Python MCP servers. Additional runtimes are WORK IN PROGRESS.

🗂️ Project Structure

.
├── docs/                       # Documentation (see INSTALL.md)
├── src/
│   └── mcpbox/
│       ├── cli/                # CLI: init, push, pull, search, scanners
│       │   ├── commands/       # CLI subcommands
│       │   └── scanners/       # SonarCloud, Bandit, ggshield, discovery
│       ├── server/             # FastAPI app + routes
│       │   ├── routes/         # servers, payment, auth
│       │   └── templates/      # Landing page
│       └── shared/             # Config, models, S3 utils
├── lambda.py                   # AWS Lambda handler (executor)
├── main.py                     # Local dev server entry
├── pyproject.toml              # Project metadata & extras
├── Dockerfile                  # Server container
├── docker-compose.yaml         # Optional local stack
└── tests/                      # Tests (placeholder)

🌐 HTTP API (Server)

Base path: /api/v1

  • Servers

    • GET /servers – list all servers
    • GET /servers/{name} – get a server by name
    • POST /servers – create a server (see schemas in mcpbox.shared.models)
  • Payment

    • POST /payment/create-order
    • POST /payment/verify-payment
    • GET /payment/payment-status/{payment_id}
  • Other

    • GET /health – config + S3 readiness
    • GET /docs – OpenAPI docs

📜 License

This project is licensed under the MIT License.

👥 Authors

About

Discover, deploy, and test MCPs in isolated sandboxes

Topics

Resources

License

Code of conduct

Contributing

Security policy

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published