Agentican Framework – OSS multi-agent for Java (github.com)

Agentican Framework, a new open-source multi-agent framework for Java, has been launched to facilitate the integration of tool-using large language model (LLM) agents into applications. This lightweight solution allows developers to define tasks in natural language, which are then transformed by the LLM-powered planner into organized workflows with specified agents and steps, eliminating the need for manual workflow definitions. With features like virtual threads for enhanced concurrency and built-in tool integrations (including over 200 SaaS applications), Agentican streamlines the development of complex AI operations while ensuring that sensitive tasks can involve a human-in-the-loop for approval. Significant for the AI/ML community, Agentican stands out due to its focus on ease of use and flexibility. It operates on Java 25+ and can integrate seamlessly with existing frameworks like Spring and Quarkus, promoting broad adoption. The framework includes capabilities such as persistent knowledge storage through a KnowledgeStore, orchestration of specialized agents that collaborate effectively, and comprehensive observability through TaskLogs, capturing every detail of the execution process. With its rich feature set and an emphasis on developer-friendly tools, Agentican is poised to empower Java developers in creating sophisticated AI-driven applications more efficiently.