Fangyuan025/Chaty

View on GitHub ↗

Private, on-device AI desktop app — GGUF (llama.cpp) & MLX models, a local coding agent, RAG knowledge base, Deep Research, vision and voice. 100% offline, no account, no telemetry. Windows & macOS.

28 ★3 forksTypeScriptUpdated 18d ago

What you need to know

A private, offline-first desktop AI app (Rust + Tauri 2) that runs open LLMs fully on-device. Includes a local coding agent, a RAG knowledge base, Deep Research, voice, and built-in MCP client support for stdio and streamable HTTP servers.

Install

Download the installer from the GitHub Releases page (Windows x64 or macOS Apple Silicon)

Usage

  • •Use the Chat/Code mode switch for a local coding agent
  • •Connect any MCP server (stdio or streamable HTTP) in the MCP settings

Key features

  • ✓100% offline LLM inference via llama.cpp (Vulkan/Metal) and native MLX on Apple Silicon
  • ✓Local coding agent with web/browser tools, precise edits and checkpoints
  • ✓Knowledge base with hybrid RAG, strict per-file citations
  • ✓Deep Research and hands-free voice
  • ✓MCP client for stdio or streamable HTTP servers

Best for

Private, off-device local AI assistant that consumes MCP servers

Caveats

  • ⚠macOS build is ad-hoc signed and not notarized (Gatekeeper warning on first open)
  • ⚠Not an MCP server itself; it is a client that connects to MCP servers
Platforms: macOS (Apple Silicon) · Windows x64

Reviewed 2026-08-11

Topics

ai-agentai-chatapple-siliconcoding-agentdesktop-appggufhuggingfacellama-cppllmlocal-llmmcpmlxmodel-context-protocolmultimodalofflineprivacyragrusttaurivoice-assistant
Stars
28★
Forks
3
Language
TypeScript
License
MIT
Created
2026-06-04
Last push
2026-08-17