# Manage and Optimize AI Agent Context with Kong Context Mesh

This talk explains Kong Context Mesh, a system for making enterprise context usable by AI agents. Chris Tam, Kong’s director of AI products, argues that the main bottleneck for agentic AI is not model intelligence but context management: agents need access to APIs, events, databases, MCP servers, and other systems, yet that context is often not agent-ready, hard to authenticate across, and difficult to maintain at scale. Context Mesh is presented as an agentic integration layer that solves those problems by converting raw sources into compact MCP-style interfaces, using progressive disclosure to reduce token bloat: agents search for relevant tools, request detailed schemas only when needed, and then execute calls through secure sandboxed code mode. Tam also describes Kong Token Vault for translating inbound corporate SSO into outbound third-party credentials without exposing secrets to agents, and a developing meta context layer that preserves agent learnings to improve future context discovery and optimization.