AWS re:Invent 2025 - From Code to Market: Build and Launch AI Agents on AWS Marketplace (ISV314)

Migrating Local AI Agents to AWS Marketplace

Building and Deploying AI Agents on AWS

• Demonstrated how to build a local AI agent using the Strang Agent open-source SDK
• Showed how to convert the local agent to be deployable on AWS Bedrock Agent Core runtime
  • Only required 4 lines of code changes to make the agent Bedrock-compatible
  • Bedrock Agent Core provides a managed, serverless runtime for AI agents
  • Handles infrastructure concerns like security, scaling, and isolation

Productizing AI Agents for AWS Marketplace

• Overview of AWS Marketplace:
  • 30,000+ transactable listings across 70+ categories
  • Millions of visitors per year, enabling global reach for products
  • Recently added categories for AI agents and tools
• AI Agent Delivery Models on AWS Marketplace:
  • Container-based agents that deploy in the buyer's AWS account
  • API-based SaaS AI agents that run in the seller's environment
  • Both models support upfront contracts and pay-as-you-go pricing
• Demonstration of Deploying a Container-based AI Agent to Marketplace:
  • Created a Marketplace listing with product details, dimensions, and pricing
  • Leveraged the Bedrock Agent Core Starter Toolkit to package the agent as a container
  • Pushed the container to a private ECR repository and linked it to the Marketplace listing
• Licensing and Metering Models:
  • Contract-based pricing with License Manager integration
    • Buyers pre-purchase licenses, which are consumed on each agent query
    • Provides visibility into remaining license entitlements
  • Pay-as-you-go pricing with Marketplace Metering Service
    • Meters agent usage in real-time, with charges based on query volume
    • Buyers only pay for what they use, without upfront license purchases

Buyer Experience and Integration

• Demonstrated the buyer experience of subscribing to and deploying the AI agent product:
  • Buyer can choose between contract-based or pay-as-you-go pricing models
  • For contract-based, the agent is deployed in the buyer's AWS account and integrates with License Manager
  • For pay-as-you-go, the agent runs in the seller's environment, and usage is metered and billed accordingly
• Showed how the buyer can interact with the deployed AI agent, either through a custom application or directly via the Bedrock Agent Core runtime

Key Takeaways

• AWS provides a comprehensive set of services and tools to help ISVs build, package, and sell AI agents on AWS Marketplace:
  • Bedrock Agent Core for managed, serverless runtime of AI agents
  • Marketplace listing options for container-based or API-based SaaS agents
  • Flexible pricing models, including upfront contracts and pay-as-you-go
  • Integrated licensing and metering to streamline the buyer experience
• Migrating a local AI agent to Bedrock Agent Core and publishing on Marketplace can be done with minimal code changes, leveraging the provided toolkits and CLI
• Selling AI agents on Marketplace enables ISVs to reach a global audience and provide a seamless, managed experience for buyers