mcp-http

README

Teamwork MCP HTTP Server

HTTP transport server for the Teamwork Model Context Protocol implementation

📖 Overview

The Teamwork MCP HTTP Server provides an HTTP interface for the Model Context Protocol, enabling secure and scalable communication between Large Language Models and Teamwork.com. This server implements the MCP specification over HTTP transport, making it suitable for production deployments and cloud environments.

🤖 What is the HTTP Server for?

This HTTP server is designed for:

• Production deployments: Scalable, stateless HTTP-based MCP communication
• Cloud environments: Easy deployment with load balancing and health checks
• Multi-client support: Handle multiple concurrent LLM connections
• Monitoring and observability: Built-in metrics, logging, and tracing

✨ Features

• HTTP Transport: API implementing the MCP specification
• Health Checks: Built-in health endpoint
• Observability: Comprehensive logging, metrics, and Datadog APM integration
• Production Ready: Designed for cloud deployment with proper error handling
• Stateless: No server-side session management for horizontal scaling

🚀 Quick Start

📋 Prerequisites

• Go 1.25 or later
• Valid Teamwork API credentials
• OAuth2 client configuration

🏃 Running the Server

# Basic HTTP server
TW_MCP_SERVER_ADDRESS=:8080 \
  go run cmd/mcp-http/main.go

# With custom configuration
TW_MCP_URL=https://my-mcp.example.com \
  TW_MCP_SERVER_ADDRESS=:8080 \
  TW_MCP_LOG_LEVEL=debug \
  go run cmd/mcp-http/main.go

🔗 Extended API Endpoints

Besides the MCP endpoints, the HTTP server provides the following extended API endpoints:

Endpoint	Method	Description
/health	GET	Health check endpoint

⚙️ Configuration

The server can be configured using the following environment variables:

Server Configuration

Variable	Description	Default	Example
TW_MCP_VERSION	Version of the MCP server	dev	v1.0.0
TW_MCP_SERVER_ADDRESS	Server bind address	:8080	:80, 0.0.0.0:80
TW_MCP_ENV	Environment the app is running in	dev	staging, production
TW_MCP_AWS_REGION	AWS region where the app is running	us-east-1	eu-west-1
TW_MCP_HAPROXY_URL	HAProxy instance URL	(empty)	https://haproxy.example.com
TW_MCP_URL	The base URL for the MCP server	https://mcp.ai.teamwork.com
TW_MCP_API_URL	The Teamwork API base URL	https://teamwork.com

Logging Configuration

Variable	Description	Default	Example
TW_MCP_LOG_FORMAT	Log output format	text	json, text
TW_MCP_LOG_LEVEL	Logging level	info	debug, warn, error, fatal
TW_MCP_SENTRY_DSN	Sentry DSN for error reporting	(empty)	https://xxx@sentry.io/xxx

Datadog APM Configuration

Variable	Description	Default	Example
DD_APM_TRACING_ENABLED	Enable Datadog APM tracing	false	true
DD_SERVICE	Service name for Datadog APM	mcp-server	teamwork-mcp
DD_AGENT_HOST	Datadog Agent host	localhost	datadog-agent
DD_TRACE_AGENT_PORT	Datadog trace agent port	8126	8126
DD_DOGSTATSD_PORT	DogStatsD agent port	8125	8125
DD_ENV	Environment for Datadog APM	(uses TW_MCP_ENV)	staging, production
DD_VERSION	Version for Datadog APM	(uses TW_MCP_VERSION)	v1.0.0

🧪 Testing

MCP HTTP CLI

The MCP HTTP CLI is a command-line tool for interacting with the MCP HTTP server. It provides a simple interface for testing API endpoints and performing common tasks.

For more information check here.

🔍 MCP Inspector

For debugging purposes, use the MCP Inspector tool:

NODE_EXTRA_CA_CERTS=letsencrypt-stg-root-x1.pem npx @modelcontextprotocol/inspector node build/index.js

[!IMPORTANT] Note: The NODE_EXTRA_CA_CERTS environment variable is required when using OAuth2 authentication with the Let's Encrypt certification authority. Download the certificate here.

🔍 Monitoring

The HTTP server provides comprehensive monitoring capabilities:

• Health Checks: /health and /ready endpoints for load balancer integration
• Structured Logging: JSON or text format with configurable log levels
• Datadog APM: Distributed tracing and performance monitoring
• Metrics: Built-in metrics for request rates, latencies, and errors