Local LLM Teaching Lab

Explore local AI with chat, an interactive tokenizer lab, token confidence, and visible agent and MCP tool traces.

Local LLM Teaching Lab makes language models easier to understand through hands-on experiments in your browser. Chat with a local model, explore how text becomes tokens, and inspect the steps behind AI responses. Why are questions like “How many r’s are in strawberry?” sometimes difficult for AI? The interactive tokenizer lab lets you compare individual letters with token fragments, inspect real token IDs, and see why tokenization can make character counting harder. With Local LLM Teaching Lab, you can: - Start immediately with a no-download teaching demo. - Download supported models and run inference locally in your browser. - Explore token probabilities, alternatives, and uncertainty with supported models. - Follow agent traces to understand how prompts, tools, and results fit together. - Experiment with teaching skills, optional MCP tools, and page context. Built for students, educators, developers, and curious learners, the extension turns abstract AI concepts into experiments you can inspect and repeat. The teaching demo uses synthetic responses and probabilities. Real models require downloads and compatible hardware. Token confidence does not guarantee correctness. Optional remote tools require consent and send their inputs to the selected service.