I see lots of posts about using 5-10 agents in coding, and how amazing LLMs are at coding, especially Claude code. I do use both claude and chatgpt (that in vscode and codex), usually start with a planning conversation, then tell them to start and have hold points to test and discuss. Neverythelss, reading some posts I feel like I am not making the most of it and achieving the full capability and throughput.

Keen to see some people share details on how they set-up their coding environment and the practical process for handling the agents.