Deploy a Minimal MCP Server on AWS Lambda — Fast & Serverless

Deploying MCP on AWS Lambda: A Minimal Serverless Guide

If you have been experimenting with MCP (Model Context Protocol) locally and want to take it to the cloud, this guide explains how to deploy a minimal MCP server using AWS Lambda and Python. This setup allows for real-world tool hosting without managing dedicated infrastructure.¹

Why AWS Lambda for MCP

Deploying your MCP server on AWS Lambda offers several advantages:

• Scalable: Automatically handles incoming requests.

• Cost-efficient: Pay only for usage.

• Secure: Integrates with IAM and AWS authorization.

• Interoperable: Easy integration with S3, DynamoDB, and other AWS services.¹

AWS MCP Architecture

Project Structure

mcp-lambda-server/
├── app.py              # MCP logic and tools
├── requirements.txt    # Python dependencies
├── template.yaml       # AWS SAM deployment config
└── event.json          # Sample test input (optional)

MCP Logic (app.py)

from aws_lambda_powertools.utilities.typing import LambdaContext
from mcp import MCPServer, tool
import json

server = MCPServer()

@tool
def hello_mcp(name: str) -> str:
    return f"Hello, {name}! Welcome to your MCP Lambda server."

@tool
def add_numbers(a: int, b: int) -> int:
    return a + b

@tool
def roll_dice() -> int:
    import random
    return random.randint(1, 6)

def lambda_handler(event: dict, context: LambdaContext) -> dict:
    return server(event)

Dependencies (requirements.txt)

aws-lambda-powertools
mcp

Install dependencies:

pip install -r requirements.txt

AWS SAM Deployment Template (template.yaml)

AWSTemplateFormatVersion: '2010-09-09'
Transform: AWS::Serverless-2016-10-31
Description: Minimal MCP Server on AWS Lambda

Resources:
  MCPFunction:
    Type: AWS::Serverless::Function
    Properties:
      Handler: app.lambda_handler
      Runtime: python3.11
      CodeUri: .
      Timeout: 10
      MemorySize: 128
      Policies: AWSLambdaBasicExecutionRole
      Events:
        MCPAPI:
          Type: Api
          Properties:
            Path: /mcp
            Method: post

Deploying to AWS

Install the AWS SAM CLI³ and run the following:

sam build
sam deploy --guided

This setup provisions:

• A Lambda function

• An HTTP endpoint (via API Gateway)

• IAM roles and security settings

Testing the Endpoint

Once deployed, test the endpoint using curl:

curl -X POST https://<your-api-id>.execute-api.<region>.amazonaws.com/mcp \
  -H "Content-Type: application/json" \
  -d '{"tool": "hello_mcp", "args": {"name": "User"}}'

Expected Response:

{
  "output": "Hello, User! Welcome to your MCP Lambda server."
}

You may replace the tool name with add_numbers or roll_dice for additional tests.

Final Thoughts

With this setup, you now have a minimal, functional MCP server that:

• Operates serverlessly on AWS

• Automatically scales with demand

• Is ready to integrate with LLMs, agents, and real-world data sources

Next steps may include:

• Adding custom tools (search, weather APIs, etc.)

• Connecting to LangChain or OpenAgents

• Integrating real-time external data sources

Acknowledgements

This guide is based on Antje Barth’s insightful talk at the MCP Summit¹², where she demonstrated how to deploy and scale MCP servers using AWS serverless technologies.

Special thanks to the AWS team and the broader MCP developer community for their continued efforts in making these tools available, scalable, and developer-friendly.

References

References