AI in Software Engineering | Riverty Tech Nights

Riverty Tech Night in collaboration with KI Park and cloud ahead

Most of what we've built over the last ~15–20 years assumes a static frame of reference: Apps talk to services over APIs. You know who's calling what, you can stick a gateway in front of it, apply some policy, and move on.
Agents don't really fit that model.

They don't just call an API and return. They decide what to do, call a tool, look at the result, call something else, maybe involve another agent, loop through an LLM a few times… all the while, doing onbehalf of the user, with permissions that aren't always obvious.
It's not always clear who is actually doing something.
It's not obvious what the full execution path looks like.

And the access control models we're used to — user → service, service → service fall part.

We'll dig into what actually breaks when you start wiring agents into real systems, especially around identity and access control. Things like:
- how (or if) you assign identity to an agent
- what it means for an agent to act "on behalf of" someone
- how you enforce policy when the flow isn't predefined
- and how you get any kind of visibility into what just happened

We'll also look at some of the patterns starting to emerge: agents talking to other agents, tools being invoked dynamically, and why a mediation layer (ie, agentgateway) must be built from the ground up with this in mind.