NanZi AI Agent Platform
An enterprise multi-agent and ChatBI platform combining multi-agent collaboration, retrieval-augmented knowledge, direct database connections, and sandboxed code execution behind role-based access control.
NanZi is an open-source, personal project offering an enterprise-oriented platform for orchestrating and executing multiple AI agents alongside a ChatBI, chat-based business intelligence, data-insight tool. Its dialogue layer supports automatic delegation to a main assistant or direct addressing of a specific expert agent, with a to-do checklist for tracking multi-step tasks and automatic scanning for available skills.
For execution, it offers five sandboxing strategies, local process, Docker, Kubernetes, a cloud sandbox service, or SSH to a remote host, with shared workspace mounting and idle-session cleanup. A persistent browser session with a live takeover panel lets a human step in during automated web tasks, and the ChatBI side manages data sources and metadata, checks for schema drift, and can self-heal a failing SQL query using a structured plan. Long-term memory persists across sessions in Redis, and the platform exposes its own MCP server, with OAuth2 authentication, alongside consuming external MCP tools. It is MIT licensed.