AI Agent Architecture — ReAct, Memory, Planning and Multi-Agent Systems · Tool-Augmented LLM — Tools, Functions and the Environment
Tool Taxonomy: Read-Only vs Effectful, Deterministic vs Stochastic
Tool-Augmented LLM — Tools, Functions and the Environment
Introduction
Not all tools are equal: a read-only tool (search, file read) can be called multiple times without worry — an effectful tool (send email, write to database) has irreversible consequences. A deterministic tool (SHA-256, calculator) always returns the same result for the same arguments — a stochastic one (LLM as sub-tool, web search) may return different results. This lesson builds a two-dimensional tool taxonomy and analyses the architectural implications of each combination for designing safe agents.