A7mdMe

MCP Production Kit

Community A7mdMe
Updated

Production-grade, type-safe Model Context Protocol (MCP) server boilerplate with dual transports, security guards, and enterprise tools.

MCP Production Kit

The enterprise-ready, type-safe TypeScript starter kit for building Model Context Protocol (MCP) servers. Built for Claude Desktop, Cursor, Antigravity, and autonomous AI agent architectures.

TypeScriptMCP SpecLicense: MIT

Free Core vs. Pro Pack Comparison

Capability Free Open-Source Core Commercial Pro Pack ($19)
MCP Spec Compliance Full (v2024-11-05 JSON-RPC 2.0) Full (v2024-11-05 JSON-RPC 2.0)
Local Stdio Transport Included (Claude / Cursor) Included
HTTP / SSE Streaming Transport Included (npm run start:sse) Included
Diagnostic Tool get_system_metrics (OS/Memory/Load) Included
Security Tool secure_http_fetch (SSRF Guard) Included
State Store Tool agent_memory_store (Namespaced KV + TTL) Included
Logging Pipeline Zero-leak stderr streaming Zero-leak stderr streaming
Glama Registry Listing Listed on Glama Registry Listed on Glama Registry
Edge Serverless Deployment - Cloudflare Workers template (global edge network)
Database Connector - PostgreSQL / Supabase with token & keyword mutation guards
Container Support Standard Dockerfile Production Docker Compose (Fly.io / Railway / AWS)
Commercial Distribution Free under MIT License Commercial License + Instant ZIP Download

Why MCP Production Kit?

Most open-source MCP examples are proof-of-concepts: they demonstrate a toy tool over local stdio, but fail when faced with real-world deployments.

MCP Production Kit solves the core engineering hurdles out of the box:

  • Dual Transport Architecture: Run locally on stdio (for Claude Desktop / Cursor) or remotely over HTTP / Server-Sent Events (SSE) using a single codebase.
  • Zero-Leak Logging: In stdio mode, stdout is strictly reserved for JSON-RPC protocol framing. Diagnostics are piped to stderr with automated secret redacting to prevent pipe corruption.
  • Hardened Security & SSRF Defense: The built-in HTTP fetch tool blocks loopback addresses, private IPv4/IPv6 ranges, and cloud metadata endpoints (169.254.169.254).
  • In-Memory Agent Persistence: Includes an isolated, namespaced key-value state store with configurable TTL for multi-step agent reasoning.
  • Production Observability: Built-in OS telemetry, memory metrics, and /health monitoring endpoint.

Architecture Overview

+-------------------------------------------------------------+
|                AI Agent (Claude / Cursor)                   |
+-------------------------------------------------------------+
                              |
                 JSON-RPC 2.0 / MCP Protocol
                              |
+-------------------------------------------------------------+
|                      Transport Layer                        |
|   +-----------------------+     +-----------------------+   |
|   |    stdio (Desktop)    |     |      HTTP / SSE       |   |
|   +-----------------------+     +-----------------------+   |
+-------------------------------------------------------------+
                              |
+-------------------------------------------------------------+
|                    MCP Production Server                    |
|   +-----------------------+     +-----------------------+   |
|   |     Zod Validator     |     |  Safe Stderr Logger   |   |
|   +-----------------------+     +-----------------------+   |
|                                                             |
|   Built-in Core Tools:                                      |
|   - get_system_metrics                                      |
|   - secure_http_fetch (SSRF Guard)                          |
|   - agent_memory_store (Namespaced KV + TTL)                |
+-------------------------------------------------------------+

Quickstart

1. Installation

git clone https://github.com/A7mdMe/mcp-production-kit.git
cd mcp-production-kit
npm install
npm run build

2. Run Verification Tests

npm test

3. Connect to Claude Desktop

Add this to your claude_desktop_config.json:

{
  "mcpServers": {
    "production-kit": {
      "command": "node",
      "args": [
        "/path/to/mcp-production-kit/dist/index.js",
        "--transport=stdio"
      ]
    }
  }
}

4. Run as Remote HTTP/SSE Server

npm run start:sse
# Listening on http://localhost:3000
# Healthcheck: http://localhost:3000/health
# SSE Stream: http://localhost:3000/sse

5. Run via Docker

docker build -t mcp-production-kit .
docker run -p 3000:3000 -e TRANSPORT=sse mcp-production-kit

Built-in Tools Specification

Tool Name Description Security Controls
get_system_metrics Real-time OS, CPU load, and heap memory breakdown. Read-only introspection.
secure_http_fetch Safe web scraping and API interaction. Blocks private IPs, AWS metadata, caps payload at 512 KB, 10s timeout.
agent_memory_store Multi-turn state management (set, get, list, delete). Isolated namespaces and automatic TTL expiration.

Commercial Pro Pack ($19 One-Time)

For production cloud deployments and enterprise teams, the Pro Pack provides ready-to-deploy extensions:

  • Cloudflare Workers Serverless Template: Low-latency edge deployment with Bearer Token auth and native streaming.
  • PostgreSQL / Supabase Connector: Read-only schema introspection and parameterized SQL queries with comment-stripping and mutation keyword defenses.
  • Production Docker Compose: Multi-stage container with non-root security, healthcheck polling, and reverse proxy setup.

Get Pro Pack ($19.00 USD) โ€” Instant Checkout on Whop (Delivery: Instant download via Whop Files app and customer billing portal upon completion).

License

MIT ยฉ 2026 Antigravity Autonomous Lab

MCP Server ยท Populars

MCP Server ยท New

    roman-rr

    Trading Signals

    17 triggers ร— 44 algorithms ร— 3 AI experts โ€” live crypto trading signals with reasoning for Claude Code, Codex, Cursor & 30+ agents. Free key; Pro adds entry, SL and TP.

    Community roman-rr
    OyadotAI

    Oya Browser

    The browser control plane for AI agents. One API over Oya Cloud, Browserbase, Steel, Anchor, Browser Use and your own Chrome, with persistent personas, CAPTCHA and MFA handling, and live human takeover.

    Community OyadotAI
    unibaseio

    Membase

    Memory layer for AI agents โ€” SDKs, CLI and MCP server for Membase

    Community unibaseio
    EternalNight996

    ๐Ÿง  memory-eternal โ€” a second brain for your AI

    Your AI forgets every session. Give it a memory you can audit: local-first, review-gated agent memory โ€” one SQLite vault for DSH, Claude Code, Codex, Cursor, OpenClaw, Hermes and MengBot (any MCP client), with an immutable audit log and a token ledger. Agents propose, humans approve. MIT, self-hosted, no cloud.

    Community EternalNight996
    quarkiverse

    Model Context Protocol Servers in Quarkus/Java

    Model Context Protocol Servers in Quarkus

    Community quarkiverse