AI agent glossary
Agentic search
Letting an agent find information the way a developer would, by listing folders, searching for text, opening files, and following links, instead of querying a vector index.
In agentic search, the agent gets ordinary search tools, such as a file lister, a text search like grep, and a file reader, and decides for itself what to look for next. It searches for a function name, opens the file, spots an import, follows it, and searches again.
Most coding agents now work this way, because code is full of exact names that text search finds reliably, and there is no index to build or keep fresh. The cost is more tool calls and more tokens per question than a single vector lookup.
Vector search still wins when people ask questions in their own words over a large body of prose. Many systems combine the two: semantic search to find a starting point, agentic search to read and verify.
Asked where refunds are calculated, a coding agent searches the repository for refund, opens the two most likely files, follows a call into a pricing module, and reports the exact function.
Learn more in the course
Related terms
Go from definitions to doing
Learn to use and build AI agents in 130 practical lessons.