Agentic Capabilities
LLMs running in loops with tools to complete multi-step tasks autonomously
What it is
Agentic AI refers to LLMs operating in an autonomous loop: receive a task, use tools to gather information, take actions, observe results, and continue until the task is complete or the model determines it's stuck. This enables tasks that can't be accomplished in a single prompt.
A coding agent, for example, might: search the codebase to understand structure, write a function, run tests, read the error output, fix the bug, and repeat until tests pass. Each step involves at least one LLM call and one or more tool invocations.
Key challenges in agentic systems: error recovery (what happens when a tool fails), context management (long tasks exceed context windows), and reliability (one bad decision early can derail the whole task).