Ah, you’re running Open Web UI but using the OpenAI API to do the inference.

You could look into hosting your own model using Ollama if you have a reasonable GPU. That was you could use something like Llama-factory to fine tune an existing available model with your own text (Nostr posts / emails / docs / etc.) and the data stays with you.

MCPs are basically tools you can give your AI. For example, a tool that interacts with your e-mail and draft e-mails for you, or a Nostr MCP, or Lightning MCP, etc.

YeH, AI loves bullet points and emoji’s :-)