Molfar System VS Code Effects

Molfar System's answer

Most AI apps give you one model and one answer. Molfar System gives you a room.

• Four participants, each on a different model and provider, answering the same question
• They can't see each other's replies — so no model echoes an opinion already voiced
• A moderator compares them, reports agreements and clashes, then drafts follow-up questions for specific participants
• A secretary keeps the project's knowledge base and writes the protocol

There's no backend: projects, chats and documents stay on the device, and requests go straight from your phone to the provider using your own API keys.

Molfar System's answer

Not because it's better at any single answer — the models are the same ones you'd use elsewhere.

It's for the cases where one answer isn't enough: comparing positions, catching what a single model missed, working through your own documents with several expert roles at once.

Practical differences:

• Free, no subscription, no in-app purchases
• No account, no sign-up, no analytics
• Nothing stored on our servers — there are none
• Works on the free tiers of all three providers

The trade-off: you manage your own API keys.

Molfar System's answer

Two groups.

People who work with documents they can't send to someone else's cloud — lawyers, analysts, auditors, consultants, researchers. Local storage isn't a feature here, it's the reason the app exists.

People curious about multi-model setups who don't want to wire up Python, a terminal and agent frameworks just to try one.

It's a semi-research project, so part of the audience is simply testing where several independent models genuinely beat one — and where the extra complexity isn't worth it.

Molfar System's answer

I'm a solo developer from Ukraine. Seven months ago I wrote the first line of this app on my phone — and never once sat down at a computer. The app, the website, the promo images: all of it was built on a phone.

That constraint shaped everything. No desktop meant no Android Studio, so the stack became Python with Kivy, packaged by python-for-android, with builds running in GitHub Actions instead of on a machine I don't have.

The idea itself came from a habit: asking the same question in three different chat windows and comparing the answers by hand. That's tedious, and worse, once you paste one model's answer into another window it starts agreeing with it. So I built the comparison into the app, and made sure the participants can't see each other.

It's a semi-research project — no monetization, no investors, no company. Part of it works well, part still needs proving.

Molfar System's answer

• Python 3.11 — the whole app, no Java or Kotlin
• Kivy 2.3 / KivyMD 1.2 — the mobile UI
• python-for-android + Buildozer — packaging into an APK/AAB
• GitHub Actions — the build machine, since there's no PC
• SQLite — local storage for projects, chats, documents and the knowledge base
• requests — direct HTTPS calls to Google Gemini, NVIDIA NIM and OpenRouter

No backend, no server framework, no cloud database — there's nothing running on our side.

Molfar System's answer

None — and by design. There's no monetization, no accounts and no analytics, so there are no customers to name and no way to know who uses the app.