Mercury enables users to build custom AI chatbots by uploading documents or scraping websites for training data. It uses OpenAI embeddings to convert text into vector representations stored in Pinecone, then retrieves relevant context to augment GPT-3/GPT-4 responses. The application supports PDF, DOCX, MD, HTML and other file formats, with a Perplexity-style chat UI that cites sources from the trained corpus.