C26.0.910 · A2A-T Telecom Agent Runtime Specialized AI agents across telecom's Business, Service, and Network layers today lack a common task language. Multi-agent interactions are unstructured, error-prone, and chat-driven. Closed-loop automation fails in production because agents cannot negotiate, delegate, or report outcomes in a standardized way. This Catalyst delivers a production-grade, trusted,secure runtime implementing A2A-T — a telecom extension of the A2A protocol — that moves from chat-based AI to protocol-driven autonomous agency. Three high-value scenarios are demonstrated end-to-end: HVS 1 — Service Fault Management (Problem Resolution). Cross-domain autonomous fault resolution spanning RAN and Transport network domains. A four-agent chain — Anomaly Detection, Problem Identification, Diagnostics, Impact Analysis, and Prioritise & Recommend — executes without human intervention, with subscriber impact surfaced through a Customer Service Digital Twin. HVS 2 — Enterprise Service Complaint Handling. An enterprise SLA breach triggers a fully autonomous complaint resolution flow across business and service layers, returning a structured outcome with audit trail and SLA credit determination. HVS 3 — Enterprise Service Fulfilment. An enterprise order triggers end-to-end autonomous service provisioning across access, transport, and service layers — coordinating across internal and external agents to compress fulfilment timelines and eliminate manual handoffs. Cross-cutting — Federated Agent Capability Discovery. A fabric-level capability demonstrating semantic, multi-ontology agent discovery — closing the gap between syntactic A2A-style registries and LLM-assisted capability matching across industry boundaries. Runs across all three HVS. The A2a-T runtime comprises three components: an Agent Registry Center with structured skills ontology and knowledge graph; an A2A-T Protocol Runtime implementing Task-T, Event-T, and Negotiation-T over a live Python/FastAPI and React/WebSocket stack; and a Trust and Audit layer covering agent onboarding and operational observability. The Catalyst targets DTW 2026 with a live, multi-vendor demonstration using real agent software and real operator data. Protocol code and reference implementations developed through this Catalyst will be contributed as open source to the Linux Foundation, making A2A-T available to the broader telecom and agentic AI community beyond TM Forum.