Show input tokens.

[SkybuckFlying]

It would be very usefull to know what kind of input tokens are being sent to google gemini.

For example an option to show them or query them after a session to get an idea of what happened and why so many input tokens were used...

[contributorai]

So maybe: CSV file called token_use.csv (or
more generically, finops.csv), showing two columns:

1. Prompt (say first X chars of the prompt)
2. Tokens used for the prompt

This could be readily used for FinOps analysis by multiplying "tokens used" by price per token to arrive at per prompt cost.

Also additional columns could display: tools used (x count), mcps used (x count) for a given prompt. This could be useful for tool and mcp developers in tuning how verbose vs effective (signal to noise) tool/mcp output is for various prompt types.