LLM Phone Home: Reliable Apps that can deliver inference from local backend

Hello all, I’m wondering what suggestions there are for an ios app that can serve an openai compatible endpoint. I am using 3sparks which works GREAT for that specific use, BUT, there is no mcp, no web search, etc. I want to show people that a local model with web search on your is very impressive, but I can’t find an app that can mimic OWUI/LMS/etc. Texting Hermes works but I was hoping to find a solution that is not using a slow agent, just calling requests from local server. So far, I tried: Apollo, Locally AI, Noema, and 3 Sparks.