Serverless Mcp

What is Serverless Mcp?

Serverless Mcp is a modern, serverless operating system for AI systems and agents, built with SST, React, and TypeScript. It provides a framework for managing Large Language Models (LLMs) and specialized AI agents through the Model Control Protocol (MCP) with a unified command system and shared operating template.

How to use Serverless Mcp?

Clone the repository, install dependencies with npm install, then start the development server with npx sst dev. Configure the test environment by copying .env.test.example to .env.test and filling in required API and AWS Cognito credentials. Run test scripts such as ./packages/brainsOS/test_scripts/mcp/test_tools.sh. Deploy to AWS with npx sst deploy --stage <stage>.

Key features of Serverless Mcp

• Visual flow editor for AI workflow design
• Unified command system for AI operations
• Secure authentication and authorization via AWS Cognito
• Real-time workflow execution
• Comprehensive audit logging

Use cases of Serverless Mcp

• Manage and orchestrate AI workflows through a visual interface
• Deploy secure, scalable AI subminds
• Perform prompt management and benchmarking
• Maintain strict data ownership and audit capabilities

FAQ from Serverless Mcp

What are the dependencies or runtime requirements?

Node.js v16 or later and an AWS account with configured credentials are required. The system uses AWS Lambda, DynamoDB, and Cognito.

Where does the data live?

Data is stored in DynamoDB, and authentication data is handled by AWS Cognito. All infrastructure runs on AWS.

How is authentication and transport handled?

Authentication is via AWS Cognito (username/password, user pool, identity pool). The API communicates over HTTPS through AWS API Gateway.

Are there any known limitations or issues?

The README notes common troubleshooting steps such as token expiration (handled automatically), rate limiting (built-in delays), and missing environment variable validation. No other explicit limits are mentioned.

Is there a test environment setup?

Yes. Create a .env.test file from the example, set secure permissions (chmod 600), and run test scripts located under packages/brainsOS/test_scripts/. Tests include interactive features (press Enter, R, Q) and show pass/fail indicators.