Serverless MCP Samples

Clone the aws-samples repository to your local machine. Each sample lives in its own subdirectory with independent deployment instructions.

git clone https://github.com/aws-samples/sample-serverless-mcp-servers.git
cd sample-serverless-mcp-servers

Review the eight available samples and select the one that matches your target architecture. For simple stateless scaling, choose stateless-mcp-on-lambda-python or stateless-mcp-on-lambda-nodejs. For session-persistent workloads choose stateful-mcp-on-ecs-python or stateful-mcp-on-ecs-nodejs. For AI agent patterns using the Strands SDK, choose strands-agent-on-lambda.

# Available samples:
# strands-agent-on-lambda           - Strands SDK agent on Lambda
# strands-agent-on-lambda-python    - Pure Python Strands agent (SAM)
# stateless-mcp-on-lambda-nodejs    - Stateless MCP on Lambda (Terraform)
# stateless-mcp-on-lambda-python    - Stateless MCP on Lambda (SAM)
# stateless-mcp-on-ecs-nodejs       - Stateless MCP on ECS (Terraform)
# stateful-mcp-on-ecs-nodejs        - Stateful MCP on ECS (Terraform)
# stateful-mcp-on-ecs-python        - Stateful MCP on ECS (SAM)
# lambda-ops-mcp-server             - Local MCP server for Lambda runtime ops

Each sample directory contains its own deployment instructions. Read the sample-specific README before deploying, as requirements and deployment commands vary significantly between samples.

cd stateless-mcp-on-lambda-python
cat README.md

For Python samples, install dependencies using pip. For Node.js samples, run npm install. SAM-based samples use 'sam build' which handles dependencies automatically during the build step.

# For Python SAM samples:
pip install -r requirements.txt
sam build

# For Node.js Terraform samples:
npm install
terraform init

Deploy using the infrastructure tool appropriate to your chosen sample. SAM samples use 'sam deploy --guided' for the first deployment. Terraform samples use 'terraform apply'. CDK samples use 'cdk deploy'.

# SAM deployment (Python samples):
sam deploy --guided

# Terraform deployment (Node.js samples):
terraform plan
terraform apply

# CDK deployment:
cdk bootstrap
cdk deploy

After deployment, AWS outputs the endpoint URL for your MCP server. Add this to your MCP client configuration as a remote server endpoint using Streamable HTTP transport.

{
  "mcpServers": {
    "serverless-mcp": {
      "url": "https://your-api-gateway-url.execute-api.us-east-1.amazonaws.com/prod/mcp",
      "transport": "http"
    }
  }
}