Optional final generation after stopAtToolNames tool call

[camidagreat]

Describe the feature

Currently, we are using stopAtToolNames inside toolUseBehavior to end an agent run when specific tools are called. We would love to be able to have optional behavior to still have the agent make one more generation after the final tool call, but prohibit additional tools from being called.

Our current workaround is using the parameters object on the tool call to define the last generation, but this makes it harder for us to keep our tools dry and reusable across agents.

[sandsc]

This is a very useful feature that lets the agent react to the stop reason when the tool stops.

[seratch]

Just to clarify, using stop_at_tool_names means your agent immediately returns result right after a specific tool call is done. This does not mean handling tool failures etc.

We don't have immediate plans to implement this because having more combined options could be confusing for developers. For now, please consider enriching your instructions (system prompt) to let an agent decide the right timing to return its final result.