ASTELD: A Six-Axis Classification Framework for Autonomous AI Agents
A recent study on arXiv has unveiled a novel classification system aimed at autonomous AI agents, named ASTELD. This system provides a clear method to evaluate and compare different AI agent platforms across six key areas: architecture pattern, security posture, tool integration model, execution paradigm, autonomy and human control levels, and deployment topology. By merging existing agent classifications with observable traits and straightforward rules for categorization, ASTELD was tested on eight representative frameworks, using OpenClaw as a detailed example. The results showed that ASTELD effectively distinguishes these platforms based on their core setups and reveals three common trends across them. You can find this paper on arXiv with the identifier 2608.05201, highlighting the need for a cohesive classification system in this field.
Key facts
- ASTELD is a six-axis classification framework for autonomous AI agents.
- The six axes are: architecture pattern, security posture, tool integration model, execution paradigm, level of autonomy and human control, and deployment topology.
- ASTELD was constructed by synthesizing prior agent taxonomies with observable platform properties and explicit category-assignment rules.
- The framework was evaluated by mapping eight representative frameworks.
- OpenClaw was used as an in-depth case study.
- The evaluation separated all eight platforms under their dominant configurations.
- Three cross-platform patterns were revealed: a security-accessibility diagonal, strong execution-architecture coupling, and a third pattern (cut off in abstract).
- The paper is available on arXiv with identifier 2608.05201.
Entities
Institutions
- arXiv