Route Your Agent Traffic through an Agent Gateway

What changes when you place a gateway between an AI agent and its models, APIs, and MCP servers? In this demo, a Fintech Support Agent investigates a disputed bank transfer using two MCP servers, a REST API, and an AI model. We run the same investigation twice: first with every dependency connected directly to the agent, and then through Postman Fabric Gateway. You'll see how Fabric Gateway centralizes credentials, routing, policies, and observability across the entire agentic system. We'll also demonstrate how progressive tool discovery replaces 104 exposed tool definitions with three search, retrieval, and execution tools—reducing total token usage from more than 200,000 tokens to approximately 40,000 while completing the same task. This is what it looks like to move from simply connecting an agent to governing those connections as one system.