OpenConduit

A cross-platform desktop AI chat application with multi-provider support and MCP tool integration.

---

Features

• Multiple AI providers — OpenAI, Anthropic (Claude), and LM Studio (local models via OpenAI-compatible API)
• MCP tool servers — connect any Model Context Protocol server over HTTP-SSE, HTTP Streamable, or stdio
• Per-tool approval — approve or deny individual tool calls before they execute, with per-server auto-approve override
• Conversation management — persistent conversations with custom titles, per-conversation provider/model selection, and context window tracking
• System prompts — global default and per-conversation system prompts
• Parameter controls — temperature, top-p, and max tokens configurable per conversation
• Markdown rendering — assistant responses rendered with syntax-highlighted code blocks
• Export — export conversations to Markdown
• Auto-updates — update checks via a Cloudflare Worker backed by GitHub Releases
• Cross-platform — macOS (Apple Silicon + Intel), Windows, and Linux

Screenshots

Coming soon.

Download

Pre-built binaries are available on the Releases page.

Platform	Format
macOS (Apple Silicon)	.zip
macOS (Intel)	.zip
Windows	Squirrel installer (.exe)
Linux	.deb / .rpm

Tech Stack

Layer	Technology
Runtime	Electron 42
Frontend	React 19, TypeScript, Tailwind CSS v4, Vite
State	Zustand v5 with persist middleware
Main-process storage	electron-store v11
Build / package	electron-forge
Update server	Cloudflare Worker

Getting Started

Prerequisites

• Node.js 20 or later
• npm 10 or later

Install & run

git clone https://github.com/OpenConduit/Client.git
cd Client
npm install
npm start

Build a distributable

# Current platform
npm run make

# macOS only
npm run make -- --platform darwin

# Windows only
npm run make -- --platform win32

Output lands in out/make/.

Configuration

On first launch, open Settings (⌘, on macOS) and add at least one AI provider:

Provider	Required fields
OpenAI	API key, default model
Anthropic	API key, default model
LM Studio	Base URL (e.g. http://localhost:1234), default model

MCP Servers

Add MCP tool servers in Settings → MCP Servers. Supported transports:

• HTTP-SSE — URL + optional headers
• HTTP Streamable — URL + optional headers
• stdio — command, args, and optional environment variables

Project Structure

src/
  main.ts                   # Electron entry point
  preload.ts                # Context bridge (window.api)
  main/
    ipc.ts                  # All IPC handlers
    providers/              # AI provider clients
    mcp/client.ts           # MCP client
    store/settings.ts       # electron-store (main process)
  renderer/
    App.tsx                 # Root component
    components/             # UI components
    hooks/                  # Chat lifecycle hooks
    stores/                 # Zustand stores (renderer)
  shared/
    types.ts                # Shared TypeScript types
worker/                     # Cloudflare Worker (update checks + feedback)

Contributing

1. Fork the repo and create a branch from main
2. Run npm run lint and ensure it passes before opening a PR
3. Fill out the pull request template — in particular the Process Boundary Check section (no Node/Electron imports in renderer files)
4. Sign the Contributor License Agreement — the CLA bot will prompt you on your first PR

License

OpenConduit is source-available under the GNU AGPL v3.0 for personal and open-source use. A separate commercial license is available for commercial deployments — contact contact@openconduit.ai for details.