What is agentic search?
Last reviewed 7 October 2026
Agentic search is a search process in which an AI model decides when to search, what to search for and whether to search again, instead of running one fixed query. The term is new and vendors define it differently. Meilisearch uses it broadly for any natural-language search experience where an LLM decides when and how to query your data.
Also called: agentic retrieval, AI agent search, LLM-driven search.
Why Agentic search matters
Search used to mean a person typing keywords and reading a list of links. Agentic search moves the typing and the reading to a model. The model can split a vague question into precise queries, notice that results are off-topic, and try different wording. For complex questions that is a real improvement; for a question about opening hours it is wasted effort.
Because the term is young, it covers several different products. Some vendors mean an LLM calling their search index as a tool. Microsoft uses the closely related name “agentic retrieval” for a pipeline in Azure AI Search that plans subqueries with an LLM and runs them in parallel. Others use it for agents that browse the open web. No standards body defines it, so compare products on what they actually do.
How it works
The common core is a loop with the model in charge. It receives a question, writes one or more queries, calls a search tool, reads the results and decides whether it has enough to answer. If not, it changes the query or searches a different source.
Microsoft’s agentic retrieval shows one concrete version. On its page read on 7 October 2026, an LLM can break a complex question into focused subqueries using chat history; the subqueries run in parallel as keyword, vector or hybrid searches; each is semantically reranked; and the results are merged with source references. Microsoft notes that this adds latency compared with a single-query pipeline.
Agentic search overlaps heavily with agentic RAG. The difference is mostly emphasis: agentic search names the search behaviour, while agentic RAG names the full pattern of retrieving and then generating an answer. Many systems do both.
Example: a support question with a typo and two asks
A customer writes: “does the pro plan inclde SSO and can I pay anualy”. A single keyword query on that text matches poorly. An agentic search step corrects the spelling and rewrites it into two queries, one about single sign-on on the Pro plan and one about annual billing.
Both queries run against the help centre. The SSO query returns a page saying the feature is on a different plan; the billing query returns the payment options page. The model answers both questions and links both pages. If one query returns nothing relevant, it says that part is unanswered instead of guessing.
How Swfte relates to it
Not a Swfte feature
Swfte does not offer agentic search as a product, and it does not sell a search index. Nearby, Swfte Cortex includes on-device knowledge bases that answer from your own files, and Swfte Studio lets you build an agent that calls a search tool in a loop, with its model chosen through Swfte Connect.
If you need a managed search index with agentic query planning, use a search vendor built for it and call it from a Studio agent through its API.
- Swfte Cortex: Answers from your own files
- Swfte Studio
- Enterprise AI search explained
Related terms
- Agentic RAG
Agentic RAG is retrieval-augmented generation in which an AI agent controls the retrieval step, deciding what to search for, which sources to use and whether to search again before it answers.
- Retrieval-augmented generation (RAG)
Retrieval-augmented generation (RAG) is a method in which a system first retrieves relevant passages from a collection of documents and then gives them to a language model as context for its answer.
- AI agent
An AI agent is a software program that uses an AI model to decide what to do next and then acts through tools to reach a goal it was given.
- Function calling
Function calling is a feature of language model APIs that lets a model return a structured request to run a function you defined, with arguments that match its schema, instead of only replying in text.
Common questions
- Is agentic search the same as AI search?
- Not exactly. AI search often means semantic or vector search that ranks results by meaning. Agentic search adds a model that decides which searches to run and whether the results answer the question. An agentic search system usually uses AI search underneath as one of its tools.
- What is the difference between agentic search and agentic RAG?
- They overlap. Agentic search describes the model planning and refining searches. Agentic RAG describes the full pattern of an agent retrieving context and then generating an answer from it. Most agentic RAG systems include agentic search, but agentic search can also return results without writing an answer.
- Is agentic search slower than normal search?
- Usually. Planning queries with a model, running several searches and judging the results takes longer than one query. Microsoft says its agentic retrieval adds latency compared with a single-query pipeline. Use it for complex questions where a better answer is worth the wait, and keep simple lookups on ordinary search.
- Does agentic search work on private data?
- Yes, if the search tools it calls can reach that data and respect permissions. The agent sees only what the tool returns, so access control belongs in the search layer, not in the prompt. Check whether a product filters results by the asking user’s permissions before you connect it to sensitive content.
Sources
Definitions on this page were read on the sources below on 7 October 2026. Where sources define the term differently, the page says so. The full glossary lists more terms.
- Meilisearch documentation overview of agentic search (read 2026-10-07)
- Microsoft Learn overview of agentic retrieval in Azure AI Search (read 2026-10-07)