Agent-Native Applications: A New Application Paradigm
4.40T1 sourcearXiv cs.MA
Source record
Published by arXiv cs.MA (T1 source). The original is at https://arxiv.org/abs/2611.23243.
Pipeline notes
The summary and note below are generated by the signal pipeline — they are Beyond Desk’s reading, not quotations from the source.
SummaryThe paper defines agent-native applications, where LLM agents supply runtime reasoning and may reshape application workflows. It models these systems as graphs of agents, tools, data sources, and human checkpoints, bounded by declarative behavioral envelopes. An orchestrator, agent mesh, trust layer, use cases, and an executable prototype illustrate operation and governance.
Why it mattersUseful to teams designing agent systems that need explicit permissions, audit trails, and bounded runtime changes, especially where human approval remains part of execution.
Cited by
No citations on record.