Launching today

Molfar System

Launching today

Run a meeting with 4 AI models on your Android phone

0 followers

Run a meeting with 4 AI models on your Android phone

0 followers

•

•

Molfar System turns a single chatbot conversation into a structured meeting. Assign up to four AI participants — each on its own model and provider (Gemini, NVIDIA, OpenRouter) — give them roles and documents, and ask your question. They answer independently, without seeing each other. A moderator compares the answers and builds one summary; a secretary keeps the knowledge base and writes the protocol. Free, no account, no server. You bring your own API keys, everything stays on your device.

Free

Launch tags:Android•Productivity•Artificial Intelligence

Launch Team / Built With

Mark by AirtopMeet Mark. Vibe automation for marketers.

Promoted

Maker

📌

Hi Product Hunt 👋 I'm a solo developer from Ukraine. Seven months ago I wrote the first line of Molfar System on my phone — and never once sat down at a computer. The app, the website, the videos: all of it was built on a phone. Hence the odd stack for Android (Python, Kivy, python-for-android, builds via GitHub Actions) — without a desktop, there wasn't much choice. What it does Instead of asking one model and getting one answer, you assemble a meeting. Up to four AI participants, each on its own model and provider (Gemini, NVIDIA, OpenRouter), each with its own role and its own documents. They answer independently and can't see each other's replies. Then a moderator compares the answers and builds one summary — shared points, disagreements, recommendations — and you can send a follow-up round of questions to specific participants. A sixth slot, the secretary, keeps the project's knowledge base and writes the meeting protocol. How it's built No backend, no accounts, no analytics, no ads, no monetization. You bring your own provider keys (all three have free tiers), and requests go straight from your phone to the provider. Projects, chats, documents and the knowledge base live on your device only. What I'm actually looking for Not customers — there's nothing to sell. This is a semi-research project: I'm testing where several independent models genuinely produce a better result, and where the extra complexity isn't worth it. Part of it works, part still needs proving. So I'm looking for people who find that question interesting. If you try it on a real task, I'd love to hear how it went. Especially when the multi-model setup gave you nothing useful. A negative result is the most valuable thing I can get right now. Limitations, upfront and honestly: knowledge base search is SQLite with LIKE, not vector-based (embeddings are on the roadmap). Free models can be slow under load. And it's Android only. Happy to answer anything about the process itself: what it's like to develop from a phone, how a meeting works under the hood, how the road to Google Play went.

Report

8d ago

Forum Threads

p/molfar-system

•

7d ago

Does asking several AI models actually beat asking one?

I spent six months building an app around that idea and I'm still not sure it's true.

The setup: four models from different providers answer the same question independently, without seeing each other's replies. A moderator then compares them and reports where they agree, where they conflict, and what to do next.

Sometimes the disagreement is the whole point one model catches a risk the others missed. Other times you get four versions of the same answer and the extra complexity bought you nothing.

So I'm curious about your experience: where did multiple independent answers genuinely win over one, and where was it just overhead?

View all