A2A‑T: Accelerating AN L4 toward the Era of Multi‑Agent Collaboration addresses one of the most persistent bottlenecks in global telecom operations: inefficient cross‑domain fault location. As network scale and complexity continue to grow, traditional O&M models struggle to manage end‑to‑end fault handling efficiently. Under its Autonomous Networks strategy, China Mobile is transforming its O&M approach to build a Level‑4, end‑to‑end Dark NOC capable of operating at massive scale while reducing manual intervention and operational burden. The solution delivers proven and significant business value following one year of commercial operation. Cross‑domain fault handling time has been reduced from hours to minutes, dispatch volume lowered by 22.4%, scheduling efficiency increased by 91%, and recovery time shortened by 16.45%. Night‑shift staffing requirements have been reduced by 36%, driving large‑scale cost reduction and efficiency improvement. These results demonstrate how intelligent automation can fundamentally improve O&M efficiency while maintaining network stability for large user populations. At the core of this innovation is the industry’s first telecom‑grade multi‑agent collective intelligence interaction protocol, A2A‑T, aligned with TMF IG1251E. The solution introduces a Supervisor Agent to enable global awareness, dynamic coordination, and optimization across domains. It standardizes multi‑vendor communication through a unified lexicon and semantic model, reduces redundant interactions using standardized APL prompt templates, and leverages LLM‑empowered O&M decision capabilities. By breaking PM, FM, and CM data silos through spatiotemporal correlation and knowledge graphs, the system enables converged, multi‑modal awareness for accurate cross‑domain root‑cause analysis, with strategy validation supported by a digital‑twin‑based simulation sandbox. Success is measured through clear operational, economic, and strategic outcomes. The solution achieves over 90% alarm auto‑diagnosis, over 90% fault localization accuracy, and a 25% reduction in MTTR. Night‑shift workload is reduced by 36%, equivalent to more than 120 digital employees, generating annual OPEX savings exceeding 20 million CNY. Beyond metrics, A2A‑T establishes a new intelligent O&M paradigm, setting a benchmark for multi‑agent collaboration and accelerating the industry’s transition toward fully autonomous, large‑scale network operations.