MASAI - Multi-Agent System AI Framework
An open-source Python framework for building multi-agent AI systems, with persistent memory, long-context management, tool integration, and both decentralized and hierarchical agent orchestration.
MASAI is a framework for building single-agent and multi-agent AI systems in Python. It supports configuring models from OpenAI, Google Gemini, and Anthropic with documented parameters, defining and using tools including Redis-backed caching, and can be used independently of LangChain.
Its documentation covers a framework overview of the architecture and design principles, a singular-agent guide for single-agent execution and memory, a multi-agent system guide for decentralized and hierarchical coordination, and an Orchestrated Multi-Agent Network (OMAN) guide aimed at enterprise-level, multi-domain systems. An AgentManager component provides the programmatic interface for managing agents.