Agentic RAG

The agent decides autonomously whether, when, and how to retrieve — issuing and reformulating queries, selecting among sources, grading the relevance of returned passages, and iterating or skipping retrieval — instead… Typical conditions for use: Some queries need external knowledge and others do not; A single retrieval pass returns noisy or insufficient context.

Maturity

Established

Cognitive loadMedium

360° overview

Agentic RAG

Six perspectives place the building block in context. The numbers show where each perspective continues in the reading path.

The building block at a glance

Agentic RAG

360°

Classification

Complexity: Medium
Impact area: Technical

Technical context

Integrations

retrieve → grade → rewrite graph (canonical).

Principles & goals

The agent decides autonomously whether, when, and how to retrieve — issuing and reformulating queries, selecting among sources, grading the relevance of returned passages, and iterating or skipping retrieval — instead of unconditionally retrieving once before generation. Named variants: Self-RAG (reflection tokens gate retrieval), Corrective RAG / CRAG (an evaluator scores evidence and triggers re-retrieval or a web fallback), GraphRAG (a knowledge graph for multi-hop questions), and RAPTOR (hierarchical tree-structured retrieval for long material).

Value stream stage: Build

Use cases & scenarios

Use cases: 3
Scenarios: 3

Compromises

Risks: 3
Best practices: 3

I/O & resources

Inputs: 3
Outputs: 3