Versions:
Kun Avatar 0.0.17, published by KunLabAI Team, is a lightweight Windows agent application that wraps the Ollama inference engine to deliver a fully on-premise conversational AI stack with enterprise-grade controls. Designed for users who need both privacy and extensibility, the program couples a main dialogue model with auxiliary models that handle context refinement, summarisation and sentiment tracking, while a recursive short- and long-term memory module continuously compresses and indexes conversation history so the agent can recall prior exchanges across sessions. Integration is provided through the Model Context Protocol (MCP), enabling SSE and Streamable HTTP endpoints that expose local tools, databases or APIs as callable functions without exposing data to the cloud. An RBAC authorisation layer and multi-tenant resource isolation make the same binary suitable for individual researchers, small teams or large organisations that must segregate projects and user data. Installation is intentionally minimal: the single Windows package ships with an embedded Ollama runtime, so the software runs out-of-the-box on any modern PC and can be accessed from other machines on the LAN. Since its first public build the Team has shipped twelve incremental versions, each tightening memory efficiency, expanding MCP compatibility and refining the administrative console. The software is available for free on get.nero.com, with downloads delivered through trusted Windows package sources such as winget, always supplying the newest release and supporting batch installation alongside other applications.
Tags: