Attributing Costs in Multi-Agent Workflows: Beyond LLM Spend

Arc_bong · reddit · 2026-08-27

As companies deploy dozens of AI agents, measuring only "LLM spend" is insufficient. Real costs include inference, tool/API calls, vector DB usage, retries, compute time, and human reviews. The post discusses the difficulty of cost attribution in multi-agent environments (where agents call other agents) and asks whether teams rely on provider bills or use infra platforms like LiteLLM, Portkey, or TrueFoundry for granular tracking.

Related event: Cost Attribution for Multi-Agent AI Workflows(2 posts)→

Original post →

More from coding & agent

coding & agent channel →