Survey maps how LLM agents moved from single tool calls to multi-tool orchestration

A new survey published in Artificial Intelligence Review traces how research on large language model agents has shifted from a narrow question—whether a model can pick and execute one correct tool call—toward a far more demanding problem: orchestrating many tools over long task trajectories. The paper, an open-access review published on 27 October 2026, is titled “The evolution of tool use in LLM agents: from single-tool call to multi-tool orchestration.”

According to the authors, tool use allows large language models to access external information, invoke software systems, and act in digital environments in ways that cannot be achieved from model parameters alone. Early research mainly studied whether a model could select and execute a correct single tool call. As agent systems have evolved, however, the focus has increasingly moved to multi-tool orchestration over long trajectories that involve intermediate state, execution feedback, changing environments, and practical constraints such as safety, cost, and verifiability.

The survey first unifies task formulations and distinguishes single-call tool use from long-horizon orchestration. It then organizes the literature around six key dimensions: inference-time planning and execution, training and trajectory construction, safety and control, efficiency under resource constraints, capability completeness in open environments, and benchmark design and evaluation.

Beyond the taxonomy, the review summarizes representative applications in software engineering, enterprise workflows, graphical user interfaces, and mobile systems.

The authors close by discussing major challenges and outlining future directions for building reliable, scalable, and verifiable multi-tool agents. As a survey, the article consolidates existing progress rather than reporting new experimental results, and its publisher notes that the version shared is a peer-reviewed accepted manuscript that may be replaced by the final Version of Record after further edits.

Subject of Research: A survey of how large language model agents have evolved from single tool calls to multi-tool orchestration over long task trajectories.

Article Title: The evolution of tool use in LLM agents: from single-tool call to multi-tool orchestration

Article References: Xu, H., Li, C., Ma, X., Ou, X., Zhang, Z., He, T., Liu, X., Wang, Z., Liang, J., Chu, Z., Liu, R., Mu, R., Tu, D., Liu, M., & Qin, B. (2026). The evolution of tool use in LLM agents: from single-tool call to multi-tool orchestration. Artificial Intelligence Review. https://doi.org/10.1007/s10462-026-11700-9

Image Credits: AI Generated

DOI: 10.1007/s10462-026-11700-9

Keywords: LLM agents, tool use, multi-tool orchestration, survey, planning and execution, safety and control, benchmarks, software engineering, large language models, Artificial Intelligence Review, evolution, tool