New AI Agent Design Framework Maps 27 Patterns Across Two Axes

Existing frameworks for LLM-based agent architectures are limited: industry guides (e.g., Anthropic, Google, LangChain) focus on execution topology—how data flows—while cognitive science surveys focus on cognitive function—what the agent does. Neither axis alone distinguishes architecturally distinct systems. For example, the same Orchestrator-Workers topology can implement Plan-and-Execute, Hierarchical Delegation, or Adversarial Verification—three patterns with fundamentally different failure modes and trade-offs.

The new framework resolves this by introducing a two-dimensional classification: (1) a Cognitive Function axis with seven categories (Context Engineering, Memory, Reasoning, Action, Reflection, Collaboration, Governance) and (2) an Execution Topology axis with six structural archetypes (Chain, Route, Parallel, Orchestrate, Loop, Hierarchy). The resulting 7x6 matrix identifies 27 named patterns, 13 with original names. The authors demonstrate orthogonality through cross-axis analysis and define eight representative patterns in detail.

Validation across four real-world domains (financial lending, legal due diligence, network operations, healthcare triage) shows the framework's descriptive coverage. Cross-domain analysis yields five empirical laws governing pattern selection—linking environmental constraints like time pressure, action authority, failure cost asymmetry, and volume to architectural choices. The framework is principled, framework-neutral, and model-agnostic, providing a common vocabulary for designing and comparing AI agent systems.