How Reactiv automates mobile commerce 80% faster with Amazon Bedrock AgentCore

Reactiv, a provider of a mobile commerce solution for Shopify merchants, has automated its app updates using Amazon Bedrock AgentCore, reducing configuration time by 80% and speeding up time-to-production by 33%. The company offers a low-code app builder and AI-powered features to help merchants launch and manage native mobile apps.

Before implementing the AI Scheduler, Reactiv had already developed a conversational AI builder that allowed merchants to modify their apps in real-time through natural language commands. However, merchants requested autonomous updates that occur on a schedule without manual intervention. To achieve this, Reactiv utilized AgentCore's three capabilities: managed agent runtime, built-in memory, and native MCP hosting.

AgentCore runs agents in Firecracker microVMs, eliminating the need for managed containers, orchestrators, and scaling policies. Built-in memory allows the platform to persist cross-session context, enabling agents to remember merchant preferences and learn from past approvals. Native MCP hosting enables the integration of Reactiv's Config MCP on Amazon Bedrock AgentCore, removing the custom authentication layer and JSON-RPC handshake code.

The AI Scheduler architecture comprises Amazon EventBridge triggering a Lambda executor, which invokes the AgentCore runtime. Three Strands agents coordinate across MCP tools, Lambda functions, AgentCore memory, and Amazon Bedrock to generate app configurations stored in Amazon DynamoDB for merchant approval. Each merchant's execution context, memory, and agent state run in their own Firecracker microVM, ensuring multi-tenant isolation.