GitHub - agentscope-ai/agentscope: Build and run agents you can see, understand and trust.
3 hours ago
- #agent-framework
- #multi-agent
- #production-ready
- AgentScope is a production-ready, easy-to-use agent framework designed for increasingly agentic LLMs, leveraging their reasoning and tool use abilities.
- Key features include simplicity (quick start with ReAct agent, tools, memory, planning, etc.), extensibility (ecosystem integrations, MCP/A2A support), and production readiness (deployment options with OTel support).
- Recent updates (2025-2026) include realtime voice agent support, database & memory compression, A2A protocol, TTS, Anthropic Agent Skill integration, and open-source projects like CoPaw and agentscope-samples.
- Installation is via pip or uv, requiring Python 3.10+, with quickstart examples for building agents, including voice-enabled and multi-agent applications.
- Supports training agents with Reinforcement Learning integration, with sample projects for math solving, Frozen Lake navigation, werewolf games, and more.
- Multi-agent orchestration is streamlined with MsgHub and pipelines for efficient message routing and conversation management.
- Includes built-in support for various capabilities like MCP, RAG, long-term memory, structured output, and deployment options (local, cloud, K8s).
- Community contributions are welcomed under Apache License 2.0, with research papers available for citation.