
Remote Calculator MCP Server
A containerized MCP server deployed on Amazon EKS that provides a simple calculator tool for adding two numbers. Demonstrates how to deploy and scale MCP servers in a Kubernetes environment with proper session management and HTTP endpoints.
README
Remote MCP Server on Kubernetes (Amazon EKS)
A containerized MCP (Model Context Protocol) server using FastMCP that provides a simple calculator tool for adding two numbers. This MCP server runs on Amazon EKS. The highlevel diagram of this project below:
Files
server.py
- FastMCP server with add_numbers toolrequirements.txt
- Python dependenciesDockerfile
- Container definitionk8s-deployment.yaml
- Kubernetes deployment and LoadBalancer servicetest_mcp_client.py
- Test Remote MCP Server
Building the Container
docker build -t mcp-calculator:latest .
Running Locally
# Run the container locally
docker run -p 8000:8000 mcp-calculator:latest
The server will be available at http://localhost:8000
Push to repo
You can push this image to either Dockerhub or an ECR repo
Deploying to Amazon EKS
Prerequisites
- AWS CLI configured with appropriate permissions
- kubectl installed
- Docker installed and running
EKS Cluster
You can use an existing EKS cluster or create a new one, it's upto you. Ensure that your terminal has access to run Kubectl on it
Deployment Steps
IMPORTANT: Change the container image to the repo URI in k8s-deployment.yaml file
- Deploy MCP Server container to EKS:
# Deploy the application
kubectl apply -f k8s-deployment.yaml
# Check deployment status
kubectl rollout status deployment/mcp-calculator
# Get service information
kubectl get services mcp-calculator-service
- Get the LoadBalancer external IP:
kubectl get services mcp-calculator-service -w
Testing the MCP Server
The server provides one tool:
add_numbers(a: float, b: float) -> float
- Adds two numbers together
Testing with curl
The MCP server is currently deployed and accessible at:
http://<insert loadbalancer url>/mcp/
Complete curl Test Workflow
Step 1: Initialize MCP Session and Get Session ID
SESSION_ID=$(curl -s -X POST \
-H "Content-Type: application/json" \
-H "Accept: application/json, text/event-stream" \
-d '{"jsonrpc":"2.0","id":1,"method":"initialize","params":{"protocolVersion":"2024-11-05","capabilities":{"roots":{"listChanged":true},"sampling":{}},"clientInfo":{"name":"curl-test","version":"1.0.0"}}}' \
http://<insert loadbalancer url>/mcp/ \
-D /dev/stderr 2>&1 | grep "mcp-session-id:" | cut -d' ' -f2 | tr -d '\r')
echo "Session ID: $SESSION_ID"
Step 2: Send Initialization Notification (Required by MCP Protocol)
curl -s -X POST \
-H "Content-Type: application/json" \
-H "Accept: application/json, text/event-stream" \
-H "mcp-session-id: $SESSION_ID" \
-d '{"jsonrpc":"2.0","method":"notifications/initialized"}' \
http://<insert loadbalancer url>/mcp/
Step 3: List Available Tools (Optional)
curl -X POST \
-H "Content-Type: application/json" \
-H "Accept: application/json, text/event-stream" \
-H "mcp-session-id: $SESSION_ID" \
-d '{"jsonrpc":"2.0","id":2,"method":"tools/list"}' \
http://<insert loadbalancer url>/mcp/
Expected Result:
event: message
data: {"jsonrpc":"2.0","id":2,"result":{"tools":[{"name":"add_numbers","description":"Add two numbers together.\n\nArgs:\n a: First number to add\n b: Second number to add\n \nReturns:\n The sum of a and b","inputSchema":{"type":"object","properties":{"a":{"type":"number"},"b":{"type":"number"}},"required":["a","b"]}}]}}
Step 4: Call the add_numbers Tool
curl -X POST \
-H "Content-Type: application/json" \
-H "Accept: application/json, text/event-stream" \
-H "mcp-session-id: $SESSION_ID" \
-d '{"jsonrpc":"2.0","id":3,"method":"tools/call","params":{"name":"add_numbers","arguments":{"a":15.5,"b":24.3}}}' \
http://<insert loadbalancer url>/mcp/
Expected Result:
event: message
data: {"jsonrpc":"2.0","id":3,"result":{"content":[{"type":"text","text":"39.8"}],"structuredContent":{"result":39.8},"isError":false}}
Testing with Python client
Replace the URL in the test_mcp_client.py, and execute it. This will act as MCP client and test the MCP server
Important Notes
- Session Timeout: Sessions expire after ~30-60 seconds due to FastMCP framework limitations
- Rapid Execution: Execute all steps quickly or use the one-liner command
- MCP Protocol: The
notifications/initialized
step is required by the MCP protocol after initialization
Cleanup
To remove from Kubernetes:
kubectl delete -f k8s-deployment.yaml
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。