BetterAIChat

A native Android AI assistant that uses your own API keys. Chat with leading language models and let the AI operate your phone: open apps, take screenshots, search the web, run shell commands, and follow reusable Skills.

BetterAIChat is a local-first agent: API keys are encrypted on-device, there is no cloud, and the AI can act on your device through function calling.

Features

Chat and models

- Multiple providers: OpenAI-compatible services (OpenAI, DeepSeek, Moonshot, Qwen, Ollama, any gateway), Anthropic Claude, and Google Gemini. - Streaming replies with Markdown rendering, reasoning stream display, and a blinking cursor. - Per-conversation Agent selection, with a built-in model catalog and server-side model fetching. - Context usage tracking, context compression, and AI-generated conversation titles. - Prompt templates for translating, summarising, polishing, writing, explaining code, and brainstorming.

Agents

- Each Agent bundles provider, API key, model, temperature, max tokens, deep reasoning, and an optional system prompt. - One-tap onboarding: paste an API key, the provider is detected from its prefix, models are fetched automatically, and the Agent is ready. - Agents can be switched per conversation and managed from Settings.

Modes

- Chat: pure conversation, no tools. - Plan: read-only analysis, write tools forbidden. - Build: every tool execution asks for confirmation. - Max: autonomous multi-step tool calls with deep reasoning.

Device control through tools

- Apps and system: open apps, adjust brightness, volume, and screen timeout, toggle flashlight, Do Not Disturb, Wi-Fi, and battery saver, read and write the clipboard, and keep the screen awake. - Media and communication: control media playback, share text, open the dialer, send and read notifications, and compose emails. - Files and documents: browse and read files in Downloads and Documents, save AI-generated content, download files from URLs, run OCR on image files, transcribe audio, and record the screen. - Automation and system: run shell commands via Shizuku, manage apps (force-stop, disable, uninstall), toggle screen state, and read the foreground app. - Information: web search across multiple engines, page reading, RSS feeds, weather, exchange rates, network latency, and device information. - UI automation: type, tap, swipe, and press Home, Back, or Recents through Accessibility, find and tap elements by their text, and read on-screen text with OCR.

Automations

- The AI can create automations triggered by time or battery level that run a sequence of tool actions unattended. - Managed and toggled from Settings.

Long-term memory

- The AI distills key user information into persistent local memory, which is injected into every request. - Memories can be managed and deleted from Settings.

Skills

- Import SKILL.md files with YAML frontmatter. - Skills can define their own tools with action types such as alarm, notification, clipboard, intent, and settings. - The app records actions into reusable Skills.

Other

- Multi-engine web search with URL deduplication and caching. - Attachments: images, PDF with on-device OCR, Word, Excel, and text files. - Voice assistant mode with a hands-free loop: the AI speaks, the microphone opens, and the spoken reply is sent automatically. - Conversation management with pinning, archiving, full-text search, starring, and Markdown export.

Requirements

- Android device - Optional: Shizuku for root-level shell commands and app management - Optional: Accessibility and Usage access for UI automation

Interface languages:

- Chinese - English

License

MIT