The most radical pattern replaces tool definitions entirely with code generation. Instead of presenting the agent with a catalog of operations, you give it two capabilities: search (explore what’s available) and execute (act on what you found).
Cloudflare’s “Code Mode” is the clearest example: 2 tools replace 2,500 endpoint definitions. The agent writes JavaScript to search the OpenAPI spec, narrows to relevant endpoints, and chains multiple API calls in a single operation.
Token footprint: 1,170,000 → ~1,000. That’s a 99.9% reduction that stays fixed regardless of API size.
Anthropic and Block’s Goose agent independently converged on the same pattern. As Cloudflare’s engineers observed: “LLMs are better at writing code to call APIs than at calling APIs directly.”
This aligns the API with how the consumer actually thinks — through code, not through enumerated menus.
Gennaro is the creator of FourWeekMBA, which reached about four million business people, comprising C-level executives, investors, analysts, product managers, and aspiring digital entrepreneurs in 2022 alone | He is also Director of Sales for a high-tech scaleup in the AI Industry | In 2012, Gennaro earned an International MBA with emphasis on Corporate Finance and Business Strategy.
