Agentic Loop

The repeated cycle an agent runs: observe the current state, think about what to do next, take an action, observe the result, repeat — until the goal is met or a stopping condition is hit.

Example: A research agent loops: search → read page → decide if it has enough info → if not, search again with a refined query.

In practice: Designing a good stopping condition is non-trivial. Without one, agents can loop indefinitely or rack up unexpected API costs.

• is a pattern within ReActReAct is a specific reasoning pattern that describes how the agent thinks and acts inside each iteration of the agentic loop.
• constrains Token BudgetA token budget sets an upper bound on how many loop iterations or LLM calls the agent can make before it must stop.