What are common cognitive architecture patterns?

ReAct (Reasoning + Acting): alternate between thinking and tool use. Plan-and-Execute: decompose goals into sub-tasks, then execute. Reflection: generate output, critique it, and revise. Tree of Thought: explore multiple reasoning paths in parallel. Memory-augmented: maintain persistent context across conversations. Most production agents combine patterns—e.g., Plan-and-Execute with Reflection for self-correction and Memory for learning from past interactions.

Cognitive Architecture

Written by Max Zeshut

Founder at Agentmelt · Last updated May 26, 2026

The structural design of an AI agent's reasoning, memory, and decision-making systems—how it perceives inputs, stores and retrieves information, plans actions, reflects on outcomes, and learns from experience. Cognitive architecture defines the agent's 'thinking' patterns: whether it uses chain-of-thought reasoning, maintains short-term and long-term memory, employs planning algorithms, or uses reflection loops to improve. The choice of cognitive architecture determines an agent's capabilities and limitations more than the underlying model—two agents using the same LLM but different cognitive architectures can perform dramatically differently on the same tasks.

Example

A research agent with a simple cognitive architecture (single prompt → single response) can answer factual questions. The same LLM in a sophisticated cognitive architecture—with planning (decompose the research question), parallel retrieval (search multiple sources simultaneously), reflection (evaluate whether findings are sufficient), and synthesis (combine results into a coherent report)—can produce analyst-quality research reports.

Frequently asked questions

What are common cognitive architecture patterns?: ReAct (Reasoning + Acting): alternate between thinking and tool use. Plan-and-Execute: decompose goals into sub-tasks, then execute. Reflection: generate output, critique it, and revise. Tree of Thought: explore multiple reasoning paths in parallel. Memory-augmented: maintain persistent context across conversations. Most production agents combine patterns—e.g., Plan-and-Execute with Reflection for self-correction and Memory for learning from past interactions.

Related glossary terms

Related niches

Back to glossary

Loading…