Agentic Retrieval-Augmented Generation: A Survey on Agentic RAG

Memory types in Agentic AI include episodic memory, semantic memory, and vector-based memory for RAG.

“Memory types include episodic memory (task-specific history) [167,168], semantic memory (long-term facts or structured data) [169,170], and vector-based memory for RAG [171, 172].”

By automating message formatting, dependency tracking, and semantic alignment, programmatic prompting ensures consistent reasoning across diverse agents in real time.

“By automating message formatting, dependency tracking, and semantic alignment, programmatic prompting prevents coordination drift and ensures consistent reasoning across diverse agents in real time [22,171].”

Structured representations such as knowledge graphs and indexing strategies appear in both RAG and agent memory communities' methods.

“For instance, structured representations such as knowledge graphs and indexing strategies appear in both communities' methods, and recent developments in agentic RAG demonstrate how autonomous retrieval mechanisms can interact with dynamic databases in ways reminiscent of agent memory architectures (Singh et al., 2025).”

Recent developments in agentic RAG demonstrate how autonomous retrieval mechanisms can interact with dynamic databases in ways reminiscent of agent memory architectures.

“For instance, structured representations such as knowledge graphs and indexing strategies appear in both communities' methods, and recent developments in agentic RAG demonstrate how autonomous retrieval mechanisms can interact with dynamic databases in ways reminiscent of agent memory architectures (Singh et al., 2025).”

By automating message formatting, dependency tracking, and semantic alignment, programmatic prompting prevents coordination drift and ensures consistent reasoning across diverse agents in real time.

“By automating message formatting, dependency tracking, and semantic alignment, programmatic prompting prevents coordination drift and ensures consistent reasoning across diverse agents in real time [21], [173].”

AI Agents benefit significantly from function calling, which extends their ability to interact with real-world systems.

“AI Agents benefit significantly from function calling, which extends their ability to interact with real-world systems [171,217].”

Agentic RAG integrates retrieval into an autonomous decision-making loop, where an LLM agent actively controls when, how, and what to retrieve.

“Agentic RAG integrates retrieval into an autonomous decision-making loop, where an LLM agent actively controls when, how, and what to retrieve (Singh et al., 2025; Sun et al., 2025e).”

Function calling extends AI Agents' ability to interact with real-world systems.

“AI Agents benefit significantly from function calling, which extends their ability to interact with real-world systems [173], [220].”

Retrieval-Augmented Generation (RAG) extends LLM capabilities by incorporating external knowledge without retraining the model.

“To overcome these challenges, Retrieval-Augmented Generation (RAG) has emerged as a promising approach to extend the capabilities of LLMs by incorporating external knowledge without retraining the model [8, 32, 48].”

By automating message formatting, dependency tracking, and semantic alignment, programmatic prompting prevents coordination drift.

“By automating message formatting, dependency tracking, and semantic alignment, programmatic prompting prevents coordination drift and ensures consistent reasoning across diverse agents in real time [22,171].”