Summary

Currently after the agent processed the request it's not visible to the A2A client how many token were consumed.

According to the A2A protocol there is no hard requirements for an A2A server to run with an LLM, I still think we should let the A2A client know the estimated tokens used for the request - it's a valuable information.

Also tool calls failures, total tool calls count, total iterations and total messages.

Acceptance Criteria

• Documentations of the A2A protocol were read and examined
• There is a clear process how this should happen between A2A server and A2A clients
• A follow up feature request ticket is created (if applicable)