Make sense of your cost

AI agent cost can feel like a mystery. Berth’s Usage screen turns it into three readable views.

Start with the three breakdowns

• By model — which models cost the most (e.g., a frontier model vs. a smaller one).
• By project — where the spend is going.
• By day — when it spiked.

Tokens tell the story

Cost follows tokens: input, output, and cache. A session that re-reads large files every turn burns input tokens; long generations burn output tokens. Berth breaks tokens down so the cause is visible, not just the total.

Mind the rate limits

Berth keeps rate-limit headroom in view, so a slowdown reads as "approaching a limit" rather than a mystery stall.