AI Glossary

Agentic RAG

Agentic AIPublished By Simon Budziak

Agentic RAG is retrieval-augmented generation where an agent controls the retrieval loop: it decides what to query, evaluates what came back, and chooses to retry, reformulate, or route to a different knowledge source before it decides the retrieved context is enough to answer.

Standard RAG runs one fixed pass, embed, fetch the nearest matches, generate, regardless of whether those matches actually contain the answer. That works for narrow, well-indexed knowledge bases, and breaks on ambiguous questions that span multiple documents. A 2025 arXiv survey on the pattern describes agentic RAG as embedding autonomous agents directly into the retrieval pipeline, using reflection, planning, and tool use to manage retrieval strategy dynamically rather than following one fixed sequence (Singh et al., arXiv:2501.09136).

Standard RAG runs one fixed pass, query, retrieve, generate. Agentic RAG runs a retrieve-evaluate loop where the agent decides to retry, reformulate, or route before generating

Standard RAG retrieves once and hopes. Agentic RAG checks its own retrieval before it lets the model answer from it.

How does the retrieve-evaluate-retry loop work?

Four steps repeat until the agent is satisfied or hits a retry limit: formulate a query, retrieve candidates, evaluate them against the question, then either generate, reformulate and retry, or escalate to a different source. This is the same plan, act, observe cycle that defines agentic AI, applied specifically to retrieval.

What architecture patterns implement this?

Corrective RAG adds a relevance check and falls back to a different source when retrieved documents score poorly. Self-RAG decides mid-generation whether it needs more evidence. Router-based RAG classifies the question first: a financial question routes to a database, a policy question to a document store, each reached through a consistent interface such as an MCP server. All three trade a cheap fixed pass for a variable number of passes that only runs as long as the question requires.

When does the added cost actually earn its keep?

It earns its cost on multi-hop questions, knowledge spread across systems with no single index, or domains where a confident wrong answer is expensive. It does not earn its cost on a narrow FAQ where one pass already answers correctly most of the time. The judgment call is the same one that governs AI agent design generally: match complexity to the task’s actual variance.

What can go wrong, and how do you catch it?

The evaluation step is itself a model’s judgment call, and a model wrong about whether it has enough evidence will retry when it should stop, or stop when it should retry, invisibly, the same failure mode covered in why self-reported confidence is not a reliable signal. Cap retries with a hard limit and a visible cost and latency budget, so a runaway loop becomes an alert instead of a silent bill, and log every retrieval attempt so a wrong answer traces back to a specific step rather than an unexplainable model error.

Frequently asked questions

How is agentic RAG different from standard RAG?

Standard RAG runs one fixed pass: embed the query, retrieve the top matches, generate an answer from them. Agentic RAG puts an agent in charge of that loop, so it can evaluate what came back, retry with a different query, or route to a different source before it generates anything.

Does agentic RAG need multiple agents?

Not necessarily. A single agent can run the evaluate-and-retry loop against one retriever. Multi-agent agentic RAG, where a router sends different question types to different specialized retrievers, is a common but optional extension once a single knowledge source stops being enough.

When is plain RAG still the better choice?

When queries are narrow, the knowledge base is small and well-indexed, and a single retrieval pass already answers correctly most of the time. The added latency and cost of an evaluate-and-retry loop is not worth it if the simple version already works.

Summarize this page with

See how this works in a real workflow