- Only works about agents, as defined in the LAS-WG Charter, are considered
- We include papers, implementations, services, and academic talks. We do not include tutorials, blog posts, and social media discussions
- Exceptions can be made on a case-by-case basis for high-profile content with widely recognized relevancy. The default on whether something fits this description is "no"
- We explicitly do not consider the following areas:
- Generic agent frameworks
- Applications of AI agents
- This includes applications of AI agents to non-AI-agent infrastructure. For example, cyber threats of AI agents are allowed, but using AI agents to identify regular cyber threats is not
- Ethics, sociology and philosophy of AI beyond infrastructural concerns
- Infrastructure for non-agentic AI (though the line can be fuzzy)
- Closed-source includes open-source SDKs for remote services (free or paid).
- A link can appear in multiple categories, though it is best to be specific about which aspect of a certain work is relevant to the area
- For example, if an agent library does both permission management and discoverability, the links should point towards the permission management and discoverability modules