OxyGent
ActiveDescription
OxyGent is an open-source multi-agent collaboration framework from JD.com, supporting flexible agent role definition, task decomposition, and collaborative orchestration for enterprise AI agent applications.
Key Features
- Modular multi-agent framework with LEGO-brick-style Oxy components for rapid assembly
- Dynamic planning with intelligent task decomposition, negotiation, and real-time adaptation
- Elastic architecture supporting any agent topology from simple ReAct to complex hybrid planning
- Built-in evaluation engines that auto-generate training data for continuous agent improvement
- Distributed scheduler enabling linear cost growth with exponential collaborative intelligence gains
- MCP integration for connecting external tools and services with Node.js runtime support
Use Cases
💡 Build enterprise-grade multi-agent systems with role-based task orchestration
💡 Deploy collaborative AI teams for domain-wide optimization and real-time decision-making
💡 Create specialized agent workflows for file operations, math computation, and time queries
💡 Scale agent systems from single-agent prototypes to distributed production deployments
💡 Implement knowledge feedback loops for continuous agent learning and improvement
Categories
Quick Start
Create a Python 3.10 environment via conda or uv. Install with `pip install oxygent`. Set environment variables for your LLM provider (DEFAULT_LLM_API_KEY, DEFAULT_LLM_BASE_URL, DEFAULT_LLM_MODEL_NAME). Write a script defining your Oxy components and run with `python demo.py`.