Research

Governed generative resource layers for agent-oriented systems.

Problem

Agent programs need access to LLMs, tools, APIs, and services without allowing those outputs to silently become beliefs, intentions, plans, or actions.

Research questions

  1. How can generated outputs be exposed to BDI agents without directly modifying agent state?
  2. What governance boundary is needed for auditable resource access?
  3. Can the same resource-layer model work across Jason, ASTRA, and related MAS frameworks?

Contribution

A Java framework boundary between agent reasoning and external generative resource use.

Academic context

Generative Layers is MSc thesis-oriented research in Advanced Software Engineering.

Dimitrios Kyriakidis — project maintainer.

Associate Professor Rem Collier — academic supervisor, Multi-Agent Systems researcher, and founder of ASTRA.