Tag
TokenCast learns composable cost representations to forecast token consumption in LLM agent execution, reducing prediction error and saving tokens in budget-controlled scenarios.