How to Reduce High Input Token Usage in AI Agent Evaluation (Oracle AI Agent Studio)?

I am currently testing an AI agent using the Evaluation and Monitoring tab in Oracle AI Agent Studio. During testing, I observed that the input tokens are significantly high compared to output tokens for certain requests.

For example:

• Input Tokens: ~98k / ~196k
• Output Tokens: ~300–400
• Status: Complete, no errors

Even for relatively simple queries like retrieving resource details or activity information, the input token count is very large.

My concerns:

• This may impact performance and cost.
• It seems like unnecessary context or payload is being passed to the model.

Additional Constraint:

• The REST API I am using does not support query parameters