El Agente Quntur: A research collaborator agent for quantum chemistry

TL;DR

El Agente Quntur is introduced, a hierarchical, multi-agent AI system designed to operate not merely as an automation tool but as a research collaborator for computational quantum chemistry, and a roadmap toward a fully autonomous end-to-end computational chemistry research agent.

Abstract

Quantum chemistry is a foundational enabling tool for the fields of chemistry, materials science, computational biology and others. Despite of its power, the practical application of quantum chemistry simulations remains in the hands of qualified experts due to methodological complexity, software heterogeneity, and the need for informed interpretation of results. To bridge the accessibility gap for these tools and expand their reach to chemists with broader backgrounds, we introduce El Agente Quntur, a hierarchical, multi-agent AI system designed to operate not merely as an automation tool but as a research collaborator for computational quantum chemistry. Quntur was designed following three main strategies: i) elimination of hard-coded procedural policies in favour of reasoning-driven decisions, ii) construction of general and composable actions that facilitate generalization and efficiency, and iii) implementation of guided deep research to integrate abstract quantum-chemical reasoning across subdisciplines and a detailed understanding of the software's internal logic and syntax. Although instantiated in ORCA, these design principles are applicable to research agents more generally and easily expandable to additional quantum chemistry packages and beyond. Quntur supports the full range of calculations available in ORCA 6.0 and reasons over software documentation and scientific literature to plan, execute, adapt, and analyze in silico chemistry experiments following best practices. We discuss the advances and current bottlenecks in agentic systems operating at the research level in computational chemistry, and outline a roadmap toward a fully autonomous end-to-end computational chemistry research agent.

Figures (15)

• Figure 1: High-level description of the multi-agentic hierarchical architecture of El Agente Quntur. Each module comprises one or more LLMs, each with the tools necessary to fulfill its role.
• Figure 2: Capabilities of El Agente Quntur and its average score as measured against the benchmark set. The benchmark is a set of 17 computational quantum chemistry exercises covering several topics from electronic and magnetic properties, thermodynamics, kinetics, and spectroscopy, as well as the various levels of theory and difficulty. The benchmark questions can be found in the \ref{['sec:benchmark']}. The rubric for each question is in the \ref{['SI:QunturEval']}\ref{['SI:QunturEval']}. Here we show the average score over 5 repetitions per question, but detailed scores for planning, geometry generation, input file generation and execution, and post-analysis are provided in Fig. \ref{['fig:benchmark']}. All benchmarks were conducted using Claude Opus 4.5 as the base language model for all agents.
• Figure 3: Simplified sketch of the case study Natural Transition Orbitals (NTOs) visualization and excited states characterization as solved by El Agente Quntur. Decisions are based on LLMs' reasoning rather than being hardcoded. Molecular structures were visualized using ChimeraX pettersen2021ucsfmengUCSFChimeraXTools2023.
• Figure 4: Development roadmap. From computational quantum chemistry research collaborator to autonomous research agent.
• Figure 5: Performance of El Agente Quntur on the benchmark set. The benchmark consists of a set of 17 computational quantum chemistry exercises covering several topics from electronic and magnetic properties, thermodynamics, kinetics, and spectroscopy, as well as the various levels of theory and difficulty. The benchmark questions are in the \ref{['sec:benchmark']}. The rubric for each question is in the \ref{['SI:QunturEval']}\ref{['SI:QunturEval']}. All benchmarks were conducted using Claude Opus 4.5 as the base language model for all agents.