# Kong AI Gateway Delivers Centralized LLM Governance and Cost Control

This talk focuses on how Kong AI Gateway helps platform teams govern developer use of AI tools, especially agentic IDEs and assistants such as Claude Code, Claude Desktop, Cursor, Copilot, and OpenCode. The speaker explains that customer demand centered less on managing fleets of agents and more on controlling developer access, visibility, and runaway token spend after public examples of AI budgets being exhausted quickly. Kong’s approach is to standardize the gateway pattern across both inference and tooling paths, including MCP servers, so organizations can centralize observability, identity, and policy enforcement. Key benefits include routing requests to providers like Anthropic or Bedrock, assigning costs to the right teams, setting fine-grained budgets, rate limiting usage, and falling back to cheaper models instead of simply cutting users off at a monthly cap. A demo shows identity-based model access with Okta, where power users can access Sonnet while standard users are limited to Haiku, with similar support noted for Entra.