Serverless MCP Framework

What is Serverless MCP Framework?

Serverless MCP Framework is an MCP server implementation built on AWS Serverless services, including Lambda, API Gateway, EventBridge, and StepFunctions. It deploys a set of Lambda functions connected to EventBridge rules for handling start, stop, and task start events, along with a StepFunction state machine for scheduled events. This framework is intended for developers who want to deploy and extend MCP servers using AWS cloud infrastructure.

How to use Serverless MCP Framework?

Use the AWS SAM CLI to deploy: run sam build followed by sam deploy --guided. Provide required parameters such as EventBusName, Project, Environment, ServerName, and authentication settings. After deployment, the MCP server endpoint is available in the stack outputs. For local development, install Node.js 22+, run npm install, configure a .env file with server and AWS settings, then start the server with npm run dev. To connect from Claude Desktop, use mcp-remote to proxy HTTP to STDIO.

Key features of Serverless MCP Framework

• Deploy MCP server using AWS Lambda, API Gateway, and EventBridge.
• Define resources, prompts, and tools via SSM parameters.
• Tools execute Lambda functions following the AWS Converse API pattern.
• Supports Streamable HTTP transport for MCP Inspector.
• Configurable authentication (token or OAuth).
• Local development mode with .env configuration.

Use cases of Serverless MCP Framework

• Deploy a scalable MCP server fully on AWS serverless infrastructure.
• Add custom resources that expose static or dynamic text content.
• Create reusable prompt templates with dynamic input interpolation using EJS.
• Implement custom tool functions as Lambda functions triggered by MCP tool calls.
• Integrate with Claude Desktop via mcp-remote HTTP-to-STDIO proxy.

FAQ from Serverless MCP Framework

How do I deploy the Serverless MCP Framework?

Deploy using AWS SAM CLI: first run sam build to build the project, then sam deploy --guided to configure parameters and create the CloudFormation stack.

How can I extend the MCP server with custom tools?

Define a tool using an SSM parameter under the prefix /<project>/<environment>/tools/<tool-name> with JSON containing name, description, and input schema. Then create a Lambda function with the naming pattern <project>-<environment>-tools-<tool-name> that returns responses following the Converse API format.

What are the runtime requirements?

For deployment, you need AWS SAM CLI and Node.js 22+. For local development, only Node.js 22+ is required.

How do I connect this server to Claude Desktop?

Add an entry to your Claude Desktop configuration that uses mcp-remote as the command, passing the deployed HTTP endpoint and the --transport http-only flag.

Where is data stored?

Resources, prompts, and tools definitions are stored in AWS Systems Manager Parameter Store (SSM) under structured prefixes. Tool execution logic is stored in Lambda functions.