Deploy sophisticated multi-agent AI applications without writing code. The recipe-based tool enables your team to quickly build and deploy AI solutions that leverage Amazon Bedrock foundation models within your own VPC.
Overview
This Guidance demonstrates how to overcome the complexity and lengthy development cycles of enterprise AI systems through a no-code multi-agent orchestration platform on AWS. The platform features an intuitive Agent Composer UI where users can build and modify agents that are automatically deployed to Amazon ECS Fargate containers. A central Supervisor Agent analyzes queries using Amazon Bedrock foundation models to route requests to specialized agents, while comprehensive security features include Amazon Cognito authentication, AWS WAF protection, and Bedrock Guardrails for safety and privacy. You can deploy sophisticated AI workflows in weeks instead of months, reducing development costs by up to 80% while gaining AI coordination capabilities that traditional chatbots cannot match.
Benefits
Accelerate AI solution development
Enhance data-driven decisions
Connect AI agents directly to your existing data platforms including Amazon Aurora, DynamoDB, and third-party systems. Your agents perform semantic search across configured data sources to deliver more relevant, contextual responses based on your organization's information.
Strengthen AI governance
Implement comprehensive security, observability, and guardrails for your AI applications. With built-in integration to Amazon Bedrock Guardrails, AWS CloudWatch, and third-party monitoring tools, you maintain control while safely scaling your AI capabilities.
How it works
These technical details feature an architecture diagram to illustrate how to effectively use this solution. The architecture diagram shows the key components and their interactions, providing an overview of the architecture's structure and functionality step-by-step.
Step 1
The user accesses the Agent Composer UI via Application Load Balancer. Amazon Cognito handles user authentication and authorization. AWS WAF provides security protection for OWASP Top 10 security risks at the entry point.
Step 2
After authentication, the Application Load Balancer routes requests to the Agent Composer API running in AWS Fargate, which retrieves existing agent configurations from AWS Systems Manager Parameter Store. Users can build new agents or modify existing ones using the Agent Composer UI also running in AWS Fargate.
Step 3
Once the user clicks Create Agent, the Agent Composer API retrieves configurations from AWS Systems Manager and triggers AWS CloudFormation to deploy a new AWS Fargate container for the agent. Once deployment completes, Amazon VPC Lattice registers the agent endpoint for multi-agent collaboration.
Step 4
Amazon Bedrock AgentCore Identity enables AI agents to securely access AWS services and third-party tools, managing both inbound authentication for user-to-agent communication and outbound authentication for agent-to-agent interactions for secure multi-agent workflows.
Step 5
The Supervisor Agent running in Amazon Fargate acts as the central orchestrator, analyzing user queries using Amazon Bedrock foundation models to determine the appropriate specialized agent to handle each request. Each specialized agent runs in dedicated Amazon Fargate containers and uses Amazon Bedrock foundation models for planning, reasoning, and response generation, with all inter-agent communication facilitated through Amazon VPC Lattice service mesh.
Step 6
If memory is enabled in the agent configuration, Amazon Bedrock AgentCore Memory or Mem0 manages session and long-term memory, providing relevant conversation context to models. The agent retrieves long-term memory from the memory provider (Amazon Bedrock AgentCore Memory or Mem0) as specified during initialization and saves conversations as short-term memory.
Step 7
If an Amazon Bedrock Knowledge Base is enabled, agents perform semantic search on configured data sources including AWS services (Amazon Aurora, Amazon DynamoDB, Amazon Bedrock Knowledge Bases) or Independent Software Vendor platforms (Snowflake, Elastic, MongoDB). The search results enhance agent responses with relevant information.
Step 8
If guardrails are enabled in the agent configuration, Amazon Bedrock Guardrails helps implement safeguards for safety, privacy, and truthfulness safeguards for the generative AI application.
Step 9
If observability is enabled, the system pushes logs, metrics, and traces to the configured platform (Amazon CloudWatch, Langfuse, Datadog, Dynatrace, Elastic, or other OpenTelemetry-compatible systems). This enables comprehensive monitoring and debugging of agent interactions and performance.
Deploy with confidence
Everything you need to launch this Guidance in your account is right here.
Let's make it happen
Ready to deploy? Review the sample code on GitHub for detailed deployment instructions to deploy as-is or customize to fit your needs.