Usage & cost

The Usage inspector summarizes token spend and estimated cost per session.

Open usage

Sidebar → Usage (chart icon) when the runtime is connected.

What you see

• Totals: input / output / cached tokens, estimated USD cost
• Group by: day, model, provider, or thread
• Cache hit rate when prompt caching applies

Display currency for labels can be USD or CNY under Settings → System (cost_currency) — backend costs remain USD-based.

Data source

Each completed turn records usage from the provider response. Estimates use bundled pricing tables; self-hosted gateways may show $0 or approximate values.

Tips

• Compare Flash vs Pro routing with model routing.
• Long LHT runs — watch context % in chat context alongside cost.
• Usage is local to your machine; not sent to zagens.com.

Related: API key · Context usage