AI Activity
Usage
See how many tokens your AI agents used and what they would cost at API prices, per day, project, session, model or agent, from each response's usage fields.
The Usage tab in AI Activity totals the tokens your agents used and estimates what they cost. The numbers come from each model response’s own usage fields, so they match what the provider counted.
Read the Usage tab
Section titled “Read the Usage tab”- Pick the period at the top right: Today, 7 days (the default) or 30 days.
- The summary line reads, for example, 14 requests · 59k new tokens · 380k read from cache · $0.62.
- A bar chart shows new tokens per day (not for Today).
- Group by switches the table between Day, Project, Session, Model and Agent.
| Column | Meaning |
|---|---|
| Requests | Model requests in the group |
| New tokens | Input, cache writes and output: what the provider processed fresh |
| Output | Tokens the model wrote |
| Read from cache | Input the provider served from its prompt cache |
| Est. cost | The estimate at API prices |
Where the numbers come from
Section titled “Where the numbers come from”- Token counts come from requests through the Watch gateway, or from an agent’s telemetry. Hooks alone don’t report tokens, and the network extension never sees them.
- Cost is Claude Code’s own estimate when its telemetry reports one, otherwise Watch’s price table, which covers Anthropic models. Models without a price show —.
- On a subscription you aren’t billed per request; the cost shows what the same tokens would cost at API prices.
The footnote under the table says which applies, for example Tokens from each response’s usage fields · cost estimated at API prices.
With nothing connected the tab says No usage yet; with no requests in the period, No model requests in this period. Pick a longer period.
Related
Section titled “Related”Connect your AI toolsThe gateway and telemetry report usage.
AI ActivityEach session's tokens and cost.