(note: on the video I only show a small part of all the models)

After 5.1.x internal versions, I finally open-sourced my Android AI app — **Kryzz AI**.

# TL;DR

Open-source Android app that gives you chat, voice, and agent workflows powered by OpenRouter (414 models), with strict local-first storage and zero telemetry. v5.1.15 source is now public on GitHub under MIT, with a debug APK ready to sideload.

# What's actually in it

* **Chat** with any of 414 OpenRouter models (per-chat reasoning controls, citations, token + cost tracking)

* **Agent workflows** (think & research / create & build — DOCX, XLSX, SQLite, code ZIPs, generated locally)

* **Voice session** with adaptive turn detection and barge-in to interrupt mid-reply while having the highest quality voice (Fishaudio)

* **Memory + library** — local Room DB for chats, folders, memory entries, skills; JSON backup/restore

* **Encrypted credential vault** — AES-GCM with a non-exportable Android Keystore key (so the keys never touch the filesystem)

# Why I built it

I wanted a single Android workspace for chat, voice, and tool-driven work where everything stays on the device. Most AI apps lock you into one vendor's UI, send analytics, and keep your history on their servers. Kryzz AI doesn't — it talks to whatever provider you point it at (OpenRouter for AI, optional Fish Audio for voice, optional Parallel for research) and stores everything locally.

# Screenshots from the v5.1.15 build

* New chat home with model picker

* Agent workflow picker (Think & Research / Create & Build)

* Voice session in Listening state

* Side drawer with chat history and folder filters

* Settings home (Appearance / AI / Voice / Memory / Tools)

* Galactic About overlay (triple-tap the logo to see it)

# Tech stack, in case anyone is curious

* Kotlin + Jetpack Compose (Material 3)

* Room (12 schema versions migrated cleanly), DataStore preferences

* OkHttp + custom SSE parser for streaming responses

* MVVM, Coroutines/Flow throughout

* Android Keystore for credential encryption

# What's next

I deliberately skipped CI, signing pipelines, and the Play Store for v1 — those are next up. Open to feedback on the model-catalog routing, the voice turn-detection, the Room schema-migration strategy, or anything else.

# Links

* **Source + README + LICENSE**: [https://github.com/kriddss3/kryzz-ai\](https://github.com/kriddss3/kryzz-ai)

* **Debug-signed APK (\~74 MB)**: [https://github.com/kriddss3/kryzz-ai/releases/tag/kryzz-AI-v1-(debug)\](https://github.com/kriddss3/kryzz-ai/releases/tag/kryzz-AI-v1-(debug))

Happy to answer anything. — Kristofers

(note: on the video I only show a small part of all the models)

After 5.1.x internal versions, I finally open-sourced my Android AI app — Kryzz AI.

TL;DR

Open-source Android app that gives you chat, voice, and agent workflows powered by OpenRouter (414 models), with strict local-first storage and zero telemetry. v5.1.15 source is now public on GitHub under MIT, with a debug APK ready to sideload.

What's actually in it

• Chat with any of 414 OpenRouter models (per-chat reasoning controls, citations, token + cost tracking)
• Agent workflows (think & research / create & build — DOCX, XLSX, SQLite, code ZIPs, generated locally)
• Voice session with adaptive turn detection and barge-in to interrupt mid-reply while having the highest quality voice (Fishaudio)
• Memory + library — local Room DB for chats, folders, memory entries, skills; JSON backup/restore
• Encrypted credential vault — AES-GCM with a non-exportable Android Keystore key (so the keys never touch the filesystem)

Why I built it

I wanted a single Android workspace for chat, voice, and tool-driven work where everything stays on the device. Most AI apps lock you into one vendor's UI, send analytics, and keep your history on their servers. Kryzz AI doesn't — it talks to whatever provider you point it at (OpenRouter for AI, optional Fish Audio for voice, optional Parallel for research) and stores everything locally.

Screenshots from the v5.1.15 build

• New chat home with model picker
• Agent workflow picker (Think & Research / Create & Build)
• Voice session in Listening state
• Side drawer with chat history and folder filters
• Settings home (Appearance / AI / Voice / Memory / Tools)
• Galactic About overlay (triple-tap the logo to see it)

Tech stack, in case anyone is curious

• Kotlin + Jetpack Compose (Material 3)
• Room (12 schema versions migrated cleanly), DataStore preferences
• OkHttp + custom SSE parser for streaming responses
• MVVM, Coroutines/Flow throughout
• Android Keystore for credential encryption

What's next

I deliberately skipped CI, signing pipelines, and the Play Store for v1 — those are next up. Open to feedback on the model-catalog routing, the voice turn-detection, the Room schema-migration strategy, or anything else.

Links

• Source + README + LICENSE: https://github.com/kriddss3/kryzz-ai
• Debug-signed APK (~74 MB): https://github.com/kriddss3/kryzz-ai/releases/tag/kryzz-AI-v1-(debug)

Happy to answer anything. — Kristofers