sample-serverless-mcp-server

What is sample-serverless-mcp-server?

sample-serverless-mcp-server is a GitHub MCP server implementation that runs on Amazon Lambda using the Streamable HTTP protocol. It is built on the official TypeScript MCP server and is intended for developers who want to deploy MCP services without maintaining persistent servers.

How to use sample-serverless-mcp-server?

Clone the repository, install dependencies, set your GitHub Personal Access Token in serverless.yml, test locally with npm sls offline, then deploy to AWS Lambda with npm sls deploy. Requires Node.js 20+, the AWS CLI, and the OSS-Serverless CLI.

Key features of sample-serverless-mcp-server

• Streamable HTTP MCP protocol support
• Elastic deployment on Amazon Lambda
• Optimized cost-performance balance
• Dynamic context priority control
• Seamless integration with existing HTTP infrastructure

Use cases of sample-serverless-mcp-server

• Deploy a GitHub MCP server without managing persistent infrastructure
• Integrate GitHub operations into MCP clients via serverless endpoints
• Leverage AWS Lambda and API Gateway for scalable MCP services
• Test MCP server behavior locally before cloud deployment

FAQ from sample-serverless-mcp-server

What does this server implement?

It implements a GitHub MCP server that uses the Streamable HTTP protocol and can be deployed on Amazon Lambda.

What are the prerequisites?

Node.js 20+, the AWS CLI (configured), and the OSS-Serverless CLI (osls) are required.

How do I provide authentication to GitHub?

Set a GITHUB_PERSONAL_ACCESS_TOKEN in the serverless.yml file. The token is stored in AWS Lambda environment variables; do not submit serverless.yml to a public repository.

Can I run the server locally for testing?

Yes. Run npm sls offline from the src/github/ directory to test locally before deploying.

What AWS services are used in the architecture?

API Gateway handles HTTP requests, Lambda executes the MCP server logic, and Streamable HTTP enables streaming responses. DynamoDB and S3 are planned for request log storage (not yet implemented).