Kong AI Gateway 2.2 Targets Enterprise Agentic Governance

The release addresses the friction between fast-moving AI architectures and traditional enterprise oversight. By decoupling AI management from standard API protocols, Kong provides a dedicated release cadence that allows teams to adopt new models without sacrificing resilience or visibility. Version 2.2 consolidates MCP server sprawl into singular routes, ensuring that agents only execute tools for which they are explicitly authorized.

Financial and platform teams gain granular control through modality-aware pricing that differentiates between text, image, and video processing, as well as cache operations. This replaces flat token-based estimates with precise cost attribution. Furthermore, Kong has integrated identity-aware policies, allowing rate limiting and analytics to follow the specific agent or user rather than a generic gateway consumer. Support for new providers including Kimi, Microsoft Foundry, and Amazon SageMaker, alongside native AWS IAM authentication for Bedrock AgentCore, positions the gateway as a central hub for diverse AI deployments. Reza Shafii, Senior Vice President of Product at Kong, noted that the update focuses on the visibility and risk management required to make agentic AI viable within large-scale corporate environments.