Chakuli - Private AI Chatbot

A private AI chatbot that runs entirely on your phone. Download an open model once, then chat offline — no account, no cloud, and your messages never leave the device.

Your conversations stay private
• Answers are generated on-device, not sent to a server
• No account or sign-up required to start chatting
• Works in airplane mode — no internet needed once a model is downloaded
• Crash reports and analytics stay off unless you turn them on

More than text
• Deep Thinking — supported models reason step by step before answering
• See & Chat — ask about a photo or image (Gemma 4)
• Listen & Chat — record audio to transcribe, translate, or reply (Gemma 4)
• Agent Skills — on-device tools that get things done; a few optional skills use the internet only if you run them

Pick the model that fits your phone
• Gemma 4 (E2B / E4B), Qwen3, Qwen3.5, SmolLM, IBM Granite, Microsoft Phi-4 Mini, DeepSeek R1, Gemma 3, and Google Gemini Nano
• Small and fast, or larger and stronger — you choose the size for your RAM
• Import any GGUF model from Hugging Face or a file already on your phone
• A built-in speed and quality benchmark shows how each model runs on your device

Fast, tunable, and honest about hardware
• GPU where it is supported, CPU elsewhere — with settings for Auto, Performance, or Battery
• Adjustable context window and faster replies (MTP) on supported models
• Works on modern 64-bit Android. About 4 GB RAM minimum; 6–8 GB or more recommended for Gemma 4 E2B

Write code, study, translate, or think out loud
Use it to draft and debug code, summarize notes, translate, plan, or talk through a hard problem — with a capable assistant that keeps the conversation on your phone.

Free, with nothing locked away
Every feature works for free. There is no paywall and no account. An optional Tip Jar supports a solo developer — and is completely optional.

Good to know
• Downloading models uses the internet; everyday chatting does not
• Model files use storage (from hundreds of MB to several GB)
• Older or very low-RAM phones can run small models, but generation is slower
• AI can make mistakes — verify important information; this is not medical, legal, or financial advice

FAQ
Is it really private? For local models, inference stays on your device.
Does it work offline? Yes — download a model once, then chat without internet.
Do I need an account? No.
Is it free? Yes. Every feature is free and the Tip Jar is optional.
Which models can I use? Gemma 4, Qwen3, Qwen3.5, SmolLM, IBM Granite, Phi-4 Mini, DeepSeek R1, Gemma 3, Gemini Nano, plus any GGUF you import.

Keep your AI on your phone. Download Chakuli and start a private, offline chat today.