CRAFT: A Domain-Agnostic Framework for Reliable Multi-Agent AI Systems

This paper presented CRAFT, a domain-agnostic framework for structuring reliable multi-agent AI applications. CRAFT organizes execution into Context Analyzer, Routing, Agent Specialization, Fusion, and Tempered Evaluation, with memory operating across the pipeline. The framework emerged from the earlier AgentHub platform proposal by refining its platform-level concepts into an explicit reusable reasoning lifecycle. The reference financial implementation demonstrates that the architecture can be executed with specialized agents, external data tools, memory, and graph-based orchestration, while a prototype evaluation test demonstrates the ability to identify a deliberately incorrect financial calculation. The next stage of the work is empirical: controlled baseline comparisons, routing evaluation, ablation studies, critic reliability measurement, multi-domain benchmarks, cost analysis, and security testing are required to establish the practical value and limitations of CRAFT.

D. B. Acharya, K. Kuppan, and B. Divya, —Agentic AI: Autonomous intelligence for complex goals—A comprehensive survey,‖ IEEE Access, vol. 13, 2025, doi: 10.1109/ACCESS.2025.3532853. C. Sun, S. Huang, and D. Pompili, —LLM-based multi-agent decision-making: Challenges and future directions,‖ IEEE Robotics and Automation Letters, vol. 10, no. 6, pp. 5681–5688, 2025, doi: 10.1109/LRA.2025.3562371. H. Apuri and C. Yepuri, —Design of multi agent autonomous workflow systems using agentic AI frameworks,‖ in Proc. Int. Conf. on AI-Driven Smart Systems and Ubiquitous Computing (ICAUC), 2026, pp. 1731–1737, doi: 10.1109/ICAUC68182.2026.11441280. M. A. Ferrag, A. Lakas, N. Tihanyi, and M. Debbah, —LLM and AI agents for autonomous systems: A survey of applications, datasets, and security challenges,‖ IEEE Open Journal of the Intelligent Transportation Systems, vol. 7, 2026, doi: 10.1109/OJITS.2026.3665677.