-
Notifications
You must be signed in to change notification settings - Fork 2.1k
Description
Is your feature request related to a problem? Please describe.
In the current A2A (Agent-to-Agent) protocol, messages are only defined with a kind of either User or Assistant. This design leads to the loss of critical information in multi-agent scenarios, such as collaborative or discussion-based workflows: specifically, which agent originated an Assistant message. This ambiguity can cause tasks executed in multi-agent environments to produce unexpected or incorrect results.
For example, in a multi-agent Werewolf game, if the messages lack a source identifier, the agents would be unable to perform analysis or make decisions based on who said what.
Describe the solution you'd like
My current workaround is to add a source field within the Message's metadata to specify the origin. However, this approach is contingent upon the A2A server-side implementation. If a given A2A server does not support parsing this field, or uses a different field name for the same purpose, this capability is lost.
Therefore, I propose that the A2A protocol should natively support the attribution of message origins. This could be achieved by introducing a dedicated source field directly into the message structure. Adopting this change would provide all A2A servers with a unified standard for identifying a message's source, thereby ensuring compatibility and reliability in multi-agent collaboration and discussion scenarios.
Describe alternatives you've considered
No response
Additional context
No response
Code of Conduct
- I agree to follow this project's Code of Conduct