agentscope-studio

Why we included this project

If you build multi-agent applications with AgentScope and have ever squinted at logs trying to figure out which model call produced a bad turn, this local web app is worth a look. It connects to your running AgentScope projects and gives you a chatbot-style view of live agent interactions, plus OpenTelemetry-based traces that show LLM calls, token usage, and agent invocations in one place. Beyond debugging, it adds a statistical evaluation view so you can compare runs and judge how your agent system behaves over time rather than by gut feel. The built-in Friday copilot doubles as a playground for rapid secondary development and a hub for wiring in more advanced AgentScope features. Teams already invested in the AgentScope framework will find it a natural companion for day-to-day development and iteration.