Skip to content
    Flow-alerta.cab
    Editorially reviewed on: 2 sources
    Your reading view

    The short answer · 26 words

    In short

    An agent loop is the repeated sequence of model decision, possible tool use, evaluation of the result and the next step up to a termination criterion.

    Flow-alerta.cab (2026): Agent Loop. https://flow-alerta.cab/en/glossar/agent-loop. Editorially reviewed on: 09 October 2026.

    How much detail would you like?

    Practical

    Core mechanism of many agentic systems; however, without limits, it can increase costs, latency and risk of error.

    Possible example: A research agent searches for a source, evaluates the result, starts a second search for gaps and ends according to fulfilled quality criteria.

    Potential applications: multi-stage research, document processing, support and automation, where interim results determine the next step.

    At a glance

    Core
    Agent runners typically perform multiple turns: the model decides on response or tool call, tool results flow back, and the cycle continues until a final result, error, or defined limit is reached.
    Use
    Core mechanism of many agentic systems; however, without limits, it can increase costs, latency and risk of error.
    Example
    Possible example: A research agent searches for a source, evaluates the result, starts a second search for gaps and ends according to fulfilled quality criteria.
    Limits
    No maximum turn number; unclear cancellation conditions; tool errors endlessly repeat; costs and runtime not monitored; side effects execute multiple times.

    In context

    Tool Use / Function Calling

    Function Calling allows a model to suggest tool calls that an application executes in a controlled manner.

    Agent Orchestration

    Coordinates agents, tools and handovers within a common flow.

    Questions about this term

    Does every agent need a loop?

    Not mandatory.

    How do you stop them?

    Through final output, error logic as well as turn, time, cost or task limits.

    All details and background

    Agent runners typically perform multiple turns: the model decides on response or tool call, tool results flow back, and the cycle continues until a final result, error, or defined limit is reached.

    ELI5 – Explained simply

    The agent does not just think once: it can act, look at the result and then choose the next step.

    Why it matters in practice

    Core mechanism of many agentic systems; however, without limits, it can increase costs, latency and risk of error.

    Example

    Possible example: A research agent searches for a source, evaluates the result, starts a second search for gaps and ends according to fulfilled quality criteria.

    Pitfalls and limits

    • No maximum turn number; unclear cancellation conditions; tool errors endlessly repeat; costs and runtime not monitored; side effects execute multiple times.

    Origin & history

    The current OpenAI Agents SDK documents the repeated sequence in the runner including max_turns.

    Use in mid-sized companies

    • Potential applications: multi-stage research, document processing, support and automation, where interim results determine the next step.

    Related terms

    Matching Flow-alerta service

    AI assistants & automationAI assistants & automation: test a suitable process with clear rights and human test points.

    Matching tool / use case

    View use casesPossible use: iterative research with termination criteria; documented example: OpenAI Agents SDK Runner.

    Sources

    Quellenprüfung: 09.10.2026. Anwendungsbeispiele sind mögliche Szenarien, keine behaupteten Kundenfälle.

    Editorially reviewed on:

    Next step

    Put knowledge into practice

    Discuss your projectView use cases