agentscope-ai/agentscope is a Python framework for building and running agents by leveraging the reasoning and tool-use capabilities of LLMs. The latest release listed in the repository is v2.0.11. However, the provided materials include no changelog for this version, so what was newly added or modified cannot be confirmed. The content below therefore describes not the release changes but the features and structure of AgentScope 2.0 as explained in the README.
The SDK provides components such as ReAct, Toolkit, Model, Context, Memory, and Workspace / Sandbox. Agent Service comes with a FastAPI backend and a prebuilt Web UI, and supports multi-tenancy and multi-session. In TeamPipeline, a leader agent can delegate tasks to member agents. This means it covers not only the operation of a single agent but also flows in which multiple agents divide roles. Python 3.11 or later is required, and the license is Apache-2.0.
TeamPipeline's delegation feature focuses on composing collaboration flows across multiple agents. However, the provided information contains no specific changes for v2.0.11. It therefore cannot be judged what this release improved over the previous version or what new features it offers. The verifiable scope extends to the documented structure of AgentScope 2.0 and the latest release number.