Per-agent daily spend limits: the architecture every AI team needs

A technical deep-dive into proxy-based budget enforcement for multi-agent LLM systems, explaining why application-level spend tracking fails and how to implement network-layer controls that make runaway costs architecturally impossible.