Is A2A Really Meant for Serverless Architecture?

[jsahil]

We have a host agent running on one Lambda, and remote agents running on separate Lambdas in our current serverless architecture. However, we are facing limitations due to cold starts — both on the host Lambda and the remote agents — which leads to significant response delays and, at times, even timeouts.

Given these challenges, we’re looking for guidance on how to achieve seamless performance with A2A in a serverless setup. Is this architecture feasible for agentic workflows, or would a server-based architecture be the more reliable option for managing these workflows?

Any insights or recommendations would be greatly appreciated!