protologue

Agentic RAG

Also called agent-based RAG, iterative retrieval.

Agentic RAG is retrieval-augmented generation in which an agent, rather than a fixed pipeline, decides whether to retrieve, what to search for, which sources or tools to use, and whether the evidence is sufficient, retrieving iteratively until it can answer.

Description

A fixed pipeline retrieves once per question, so it fails on multi-hop questions and on queries that need reformulation. Agentic RAG handles them by planning, issuing follow-up searches, and reflecting on results, at the cost of more model calls, higher latency, and less predictable behavior. Self-RAG, corrective RAG, and active retrieval are earlier steps in the same direction.

Sources

  1. Singh et al. (2025). Agentic Retrieval-Augmented Generation: A Survey on Agentic RAG.

Cite this entry

Protologue. (2026). Agentic RAG. In Protologue: A Taxonomy of Prompting and LLM Techniques (v1.4.0, PTL-0134). https://protologue.com/t/agentic-rag/

BibTeX
@misc{protologue_agentic_rag,
  title = {Agentic RAG},
  author = {{Protologue}},
  year = {2026},
  howpublished = {Protologue: A Taxonomy of Prompting and LLM Techniques, v1.4.0},
  note = {Entry PTL-0134},
  url = {https://protologue.com/t/agentic-rag/}
}

Markdown JSON