AutoGen AgentChat makes team control an explicit runtime surface

AutoGen is useful because it treats multi-agent collaboration as a runtime surface with named controls instead of a loose prompt pattern. The AgentChat guide documents teams as groups of agents working toward a common goal, with presets such as `RoundRobinGroupChat`, `SelectorGroupChat`, `MagenticOneGroupChat`, and `Swarm`; it also points to GraphFlow workflows, memory and RAG, human-in-the-loop, and Studio. The teams tutorial emphasizes streaming observation through `run_stream()`, `TaskResult` stop reasons, reset behavior, and external termination that lets the current agent finish its turn before the team stops so shared state stays consistent. For agent builders, the practical signal is to evaluate team runtimes by whether speaker choice, handoff transitions, workflow graphs, memory boundaries, human review, streaming telemetry, stop conditions, and reset/resume semantics are explicit enough to debug and operate.