Agentic Systems Vocabulary

an AI system capable of autonomous action to achieve goals

“The agent decomposed the goal into steps and executed them.”

Origin: Latin agere `to do, act`

the ability of an AI to creatively utilize external functions or APIs

“Tool use allows the LLM to browse the web or calculate math.”

Origin: Old English tol `instrument`

a structured way for LLMs to invoke code functions with specific arguments

“The model used function calling to get the current weather.”

Origin: Latin functio `performance` + calling

Reasoning + Acting; a pattern where models think before executing actions

“ReAct prompting helps agents solve multi-step problems reliably.”

Origin: Acronym: Reasoning and Acting (Yao et al., 2022)

the ability to formulate a sequence of actions to achieve a future goal

“Strategic planning lets the agent handle complex, long-horizon tasks.”

Origin: Latin planum `level ground` (drawing on a flat surface)

the process of analyzing past actions to improve future performance

“Through reflection, the agent corrected its own coding error.”

Origin: Latin reflectere `to bend back`

systems for storing and retrieving information over time

“Long-term memory allows the agent to recall user preferences.”

Origin: Latin memoria `mindful`

systems involving multiple interacting agents with distinct roles

“A multi-agent team of research, writing, and editing agents wrote the paper.”

Origin: Latin multi- `many` + agent