How MoEngage turned over 400 APIs into agent tools with AgentCore Gateway
aws ·
MoEngage successfully integrated over 400 internal REST APIs into AI agent tools using Amazon Bedrock AgentCore Gateway, addressing significant challenges in security and scalability. The company rejected the traditional approach of deploying individual Model Context Protocol (MCP) servers for each team, which would have required extensive networking and maintenance across seven AWS Regions. Instead, they established a principle where product teams provide OpenAPI specifications while the AI Platform team manages the MCP layer end-to-end. This architecture ensures that private APIs remain within virtual private clouds without public endpoints, keeping identity context out of the LLM's reach. By utilizing Cedar policies for server-side authorization, MoEngage filters tool visibility per agent before serialization, preventing context window overflow from large catalogs. The solution also employs semantic search to allow agents to discover tools by intent rather than name, reducing fragility during API updates. This centralized approach saved an estimated six months of engineering effort and reduced new API onboarding time from days to hours.
aws