atulpandey84/gcp-multi-agent-team ? reverse-engineered prompt

Build me a Python app for a GCP multi agent engineering team that can coordinate different specialist agents to help with architecture, reasoning, coding, and fast routine tasks. I want it to feel like a small internal assistant platform, with a safe default setup so it does not make any GCP changes or deploy anything unless a human approves it first.

Please include a simple local monitoring UI so I can open it in the browser, connect with an API key, and see or control what the agents are doing. Set it up so it runs locally with a virtual environment, has tests, and can be started from the command line on both macOS or Linux and Windows. Use current NVIDIA model names and docs if you need to confirm anything. Also make sure the project is organized cleanly enough that it could grow into a real multi agent engineering workflow later.