Death by Tool Calls: Using CodeGen Agents to Solve the Biggest Problem with Agentic AI

AI agents are all the rage right now, and companies are rushing to find creative ways to put them to work. Especially compelling is multi-agent orchestration, in which swarms of agents, each armed with a particular set of skills, collaborate as a team to accomplish complex tasks.

What if you were told that everything you learned about agentic AI was either wrong or incomplete? That many agentic apps won’t scale, are prone to costly “token spikes,” and can fail outright by blowing up the LLM’s context window? What if it turned out that MCP makes matters worse, not better? The solution is CodeGen agents, which bypass the tools protocol and generate code that calls tools directly. Learn what CodeGen agents are, how to write them, and how to use them to deliver cutting-edge agentic apps. It might just change the way you think about agentic AI.