MCP Agent Fine-Tuning Workshop

Learn to fine-tune AI agents using Model Context Protocol (MCP) servers in this hands-on workshop that guides you through generating high-quality reasoning traces and training a Qwen3 model. Set up and run an agent with access to MCP servers, including a Playwright browser automation tool, while exploring how to integrate additional servers for expanded functionality. Generate sophisticated reasoning traces through an OpenAI-style endpoint, capturing the decision-making processes of your agent as it interacts with web browsers and other tools. Master the techniques for saving both individual tool interactions and complex multi-turn conversation traces that demonstrate advanced reasoning patterns. Apply fine-tuning methodologies using Unsloth to train a Qwen3 model on your collected reasoning traces, transforming raw agent interactions into a specialized model. Deploy and test your fine-tuned model to evaluate its performance and understand how the training process has enhanced its reasoning capabilities with MCP server integration.