About Sprapp

Own your AI. Build it, run it, sell it.

Our mission: AI shouldn't be something you rent forever. Sprapp builds its own tiny language models and the CPU engine that runs them — so you can train a custom agent, run it cheap and private, and own what you build.

The Problem

Today's AI is rented from a handful of giants. You pay per token, forever; you're locked to one vendor's roadmap and pricing; and you can't truly fine-tune or own a closed model. The more you use it, the more you pay — and you still own nothing.

For a quick question, that's fine. But if you're building a product, a classroom tool, or an agent you depend on, renting intelligence by the token is a tax that never goes away — and a model you can't control is a foundation you don't own.

The Solution

Sprapp owns the whole stack — the model and the engine:

Our own models: a 1.58-bit ternary LLM family — eeny, meeny, and miny (360M, 211 MB) — built from scratch and small enough to run on a plain CPU.
Our own engine: serve-native, a Rust inference engine with hand-written AVX2 SIMD, serving miny at up to 58 tok/s with no GPU — OpenAI-compatible, with live LoRA adapter hot-swap.
Train your own: give an agent a persona and a few examples, fine-tune a LoRA adapter in minutes, and run it private or publish it to earn.

The result: fast, cheap, private AI on models you actually own — with a flat cost that falls the more you use it.

How It Works

Public chat runs entirely on-device in your browser via WASM — your conversations never leave your device, and inference costs $0 at any scale. For hosted use, your trained adapter hot-swaps into the serve-native engine on commodity CPU.

Train an agent in Studio, chat with it in the Playground, screen prompts with Sprappy Filter, and publish to the Marketplace — all on the models we built.

Key Features

Fast CPU inference: miny at up to 58 tok/s on a plain CPU, GPU optional
Train-your-own: persona + examples → a real LoRA adapter in minutes, with a report card
On-device privacy: public chat runs in the browser; conversations never leave your device
Marketplace: publish the agent you trained and earn — own it, subscribe, or pay per query
Sprappy Filter: pre-LLM threat scoring across 25 categories in under a millisecond
Live, cited web: agents ground answers in real-time web results via Exa

The Name

Sprapp is about wrapping real ownership around AI — your model, your engine, your agent. Build it, run it, sell it.

Community

We welcome feedback, feature requests, and bug reports. Get in touch via our contact page or email us directly.

The Team

Sprapp is built by BizTransit Sdn Bhd, a technology company based in Kuala Lumpur, Malaysia. We are a small, focused team building tiny, ownable AI.

BizTransit Sdn Bhd
Level 28, Lingkaran Syed Putra
Mid Valley City, Kuala Lumpur 59200, Malaysia

Get in Touch

We'd love to hear from you. Whether you have feedback, questions, partnership ideas, or just want to say hello:

Email: hello@sprapp.com
GitHub: github.com/sprapp/support
Full contact details: Contact page