CopilotKit launches CLiF: a platform that transforms agent and user interactions into skills and training data

CopilotKit, the company behind the open-source CopilotKit framework and AG-UI protocol, has released CLiF, a continual learning platform that converts agent traces and user actions into skills, evaluations, and fine-tuning data. The platform captures full-stack trajectories through AG-UI, the open protocol connecting agents to user-facing applications. CLiF securely records both agent and user interactions, enabling teams to extract insights, build skills, and export data for training and evaluation. CopilotKit is open source with over 38,000 GitHub stars. CLiF operates through four stages: capturing trajectories, identifying patterns, reviewing insights, and delivering skills. Teams can manually or automatically trigger analysis, which surfaces recurring user corrections and friction points. Skills are proposed as SKILL.md files and require human approval before being used by agents. The platform supports multiple agent frameworks, including LangGraph, Mastra, and Google ADK. For technical leaders, CLiF addresses challenges in scaling AI agents by providing self-hosted options on Kubernetes, AWS, or managed cloud services. Trajectories, skills, and insights remain within the company’s network, ensuring data security. CLiF also generates eval cases and fine-tuning datasets, helping teams validate and improve agent performance.