Getting an agent working locally has become ridiculously easy. The moment you want someone else to depend on it, everything changes. You need environments, secrets, permissions, monitoring, evaluations, versioning, rollback and some way to know whether the new version is actually better.
It feels strange that the development side of agents has matured so quickly while the production workflow still feels fragmented. Frameworks can get you to a working agent, but what happens between "works on my machine" and "this handles a business process every day"?
[link] [comments]