sample-agentic-ai-demos

This MCP server bundle provides a collection of Agentic AI demos for AWS using the Model Context Protocol (MCP). It includes examples of how to connect an Agentic AI workflow to Bedrock via MCP SSE and demonstrate interactions across multiple language implementations (TypeScript, Python, Java, Kotlin) and deployment environments (local, ECS). The server acts as a central MCP endpoint that other clients can connect to over MCP SSE or MCP stdio, enabling you to experiment with Bedrock-based agents, RAG workflows, and service orchestration. To use it, run the provided Docker image to start the MCP server, then deploy the demo clients or agents that target the MCP endpoint. Explore the modules to see concrete samples such as a FastAPI/MCP SSE server, a Spring AI MCP server on ECS, and Python-based MCP demos that integrate with Anthropic Bedrock or Bedrock via the Converse client.

Prerequisites:

• Docker installed and running on your host
• Access to AWS credentials configured (for Bedrock and AWS services used by the demos)

Installation steps:

1. Ensure Docker is up and running:
   • On Windows/macOS: use Docker Desktop
   • On Linux: start docker service (e.g., sudo systemctl start docker)
2. Pull and run the MCP server image: docker pull aws-samples/sample-agentic-ai-demos docker run -i aws-samples/sample-agentic-ai-demos
3. Verify the server is reachable (check logs for MCP handshake messages or access the exposed endpoint if enabled in the container).
4. If you need to customize AWS credentials, set environment variables as described in the mcp_config section before running the container, or mount a credentials file into the container as appropriate.

Notes and tips:

• The included demos cover multiple language implementations and deployment options; pick the module that matches your environment (local Python, TypeScript, Spring/Kotlin on ECS, etc.).
• If you expose the MCP endpoint publicly (e.g., via ECS or Load Balancer), ensure proper security controls are in place for MCP connections and AWS credentials.
• Environment variables for AWS access should be kept secure; prefer IAM roles or secret managers where possible in production.
• Logs from the Docker container can help diagnose MCP connection issues or client handshakes; monitor container logs for MCP protocol events.
• If you modify or extend demos, maintain compatibility with the MCP SSE protocol and ensure clients point to the correct server endpoint.