Federated Gateway Layer for Enterprise Conversational AI Agents - Supporting Agentic Enterprise AI Through a Unified Discovery and Messaging Infrastructure

Abstract

This thesis presents a federated gateway layer designed to improve how users interact with distributed enterprise conversational AI agents. In large organizations, services are often fragmented, making it difficult to discover and access relevant information. The proposed system introduces a lightweight API gateway that enables unified service discovery and message routing without centralizing domain systems. The gateway uses metadata-driven filtering and deterministic routing to identify suitable agents, ensuring predictable and explainable behavior. It is integrated with conversational interfaces through the Model Context Protocol (MCP), allowing AI assistants to discover and communicate with services via structured tool calls. A registry-based design supports flexible search, including fuzzy matching, while sensitive data is securely managed through encryption. The implemented system demonstrates that a gateway-based approach can reduce complexity, improve discoverability, and provide a consistent interaction model in enterprise environments