OrchestraAI: A Multi-Agent Generative AI for Software Development
Abstract
This paper presents OrchestraAI, a multi-agent AI assistant designed to support professional software developers by combining conversational interaction with autonomous execution of development tasks, including code generation and Git version control. Unlike typical AI chatbots, OrchestraAI actively performs actions such as creating and modifying code files within a customizable local environment, preserving privacy and offering flexibility through the integration of open-source and external language models. The paper details the system’s modular architecture, which leverages LangChain and pluggable Large Language Models (LLMs), and reports qualitative results from initial evaluations. The findings indicate a strong potential for productivity improvement, along with identified limitations and proposed future improvements.References
Y. Cao, S. Li, Y. Liu, Z. Yan, Y. Dai, P. S. Yu, and L. Sun. A comprehensive survey of ai-generated content (aigc): A history of generative ai from gan to chatgpt. J. ACM, 37 (4):111:1–111:44, 2024.
M. e. a. Chen. Evaluating large language models trained on code. arXiv, 2021. DOI: 10.48550/arXiv.2107.03374. URL [link]. Trabalho seminal (Codex) demonstrando geração de código funcional a partir de descrições em linguagem natural e resultados em benchmarks como HumanEval.
V. S. Narajala and O. Narayan. Securing agentic ai: A comprehensive threat model and mitigation framework for generative ai agents. arXiv preprint arXiv:2504.19956, 2025.
L. Ning, Z. Liang, Z. Jiang, H. Qu, Y. Ding, W. Fan, X. yong Wei, S. Lin, H. Liu, P. S. Yu, and Q. Li. A survey of webagents: Towards next-generation ai agents for web automation with large foundation models. arXiv preprint arXiv:2505.23350, 2025.
L. Ouyang, J. Wu, X. Jiang, D. Almeida, C. L. Beam, C. Keller, A. Madan, W. Agarwal, J. Schulman, M. Chen, T. Brown, et al. Training language models to follow instructions with human feedback. In Advances in Neural Information Processing Systems, volume 35, 2022.
R. Sapkota, K. I. Roumeliotis, and M. Karkee. Ai agents vs. agentic ai: A conceptual taxonomy, applications and challenges. arXiv preprint arXiv:2505.10468, 2025.
J. Xu, Z. Li, W. Chen, Q. Wang, X. Gao, Q. Cai, and Z. Ling. On-device language models: A comprehensive review. arXiv, 2024. DOI: 10.48550/arXiv.2409.00088. URL [link].
M. e. a. Chen. Evaluating large language models trained on code. arXiv, 2021. DOI: 10.48550/arXiv.2107.03374. URL [link]. Trabalho seminal (Codex) demonstrando geração de código funcional a partir de descrições em linguagem natural e resultados em benchmarks como HumanEval.
V. S. Narajala and O. Narayan. Securing agentic ai: A comprehensive threat model and mitigation framework for generative ai agents. arXiv preprint arXiv:2504.19956, 2025.
L. Ning, Z. Liang, Z. Jiang, H. Qu, Y. Ding, W. Fan, X. yong Wei, S. Lin, H. Liu, P. S. Yu, and Q. Li. A survey of webagents: Towards next-generation ai agents for web automation with large foundation models. arXiv preprint arXiv:2505.23350, 2025.
L. Ouyang, J. Wu, X. Jiang, D. Almeida, C. L. Beam, C. Keller, A. Madan, W. Agarwal, J. Schulman, M. Chen, T. Brown, et al. Training language models to follow instructions with human feedback. In Advances in Neural Information Processing Systems, volume 35, 2022.
R. Sapkota, K. I. Roumeliotis, and M. Karkee. Ai agents vs. agentic ai: A conceptual taxonomy, applications and challenges. arXiv preprint arXiv:2505.10468, 2025.
J. Xu, Z. Li, W. Chen, Q. Wang, X. Gao, Q. Cai, and Z. Ling. On-device language models: A comprehensive review. arXiv, 2024. DOI: 10.48550/arXiv.2409.00088. URL [link].
Published
2025-09-22
How to Cite
HOLANDA, João Luiz Bione da Silva; SOUZA, Thiago Silva de.
OrchestraAI: A Multi-Agent Generative AI for Software Development. In: WORKSHOP ON BOTS IN SOFTWARE ENGINEERING (WBOTS), 2. , 2025, Recife/PE.
Anais [...].
Porto Alegre: Sociedade Brasileira de Computação,
2025
.
p. 47-56.
DOI: https://doi.org/10.5753/wbots.2025.14664.
