# An AI Registry Lets Developers Instantly See Which APIs, Models and Tools They Can Use

This talk explains how enterprises can operationalize agentic enablement so developers and AI agents can quickly know which APIs, MCP servers, models, and tools they are allowed to use in the right context. Miko, a product manager at Kong, frames the problem through a new developer who needs a simple answer to “What can I use?” without searching wikis or asking multiple teams, then argues that adoption is already ahead of governance, citing figures that 86% of organizations and 91% of enterprises already have AI agents publishing code, while Gartner data suggests Fortune 500 companies may grow from about 15 agents on average last year to more than 150,000 by 2028, with only 13% feeling they have the right controls. Kong’s answer is an AI registry: a discovery and consumption layer that centralizes assets, requires human approval, scopes access at the right audience and granularity, supports different interfaces for developers, IDEs, and agents, separates discovery from invocation, and stays continuously groomed so results remain current and trustworthy.