Agent Skills

SimpleChatJS

A lightweight AI chat application with MCP support, built with pure JavaScript and Node.js.

Install

Install and configure the MCP from https://github.com/Tomobobo710/SimpleChatJS now. Follow the repository's installation instructions, ask me for anything you can't complete yourself, and verify its tools load.
README

SimpleChatJS

An Electron desktop AI chat application with MCP (Model Context Protocol) support.

Tech Stack

  • Frontend: Vanilla JS, HTML, CSS (no frameworks)
  • Backend: Node.js + Express
  • Desktop: Electron
  • Database: SQLite (better-sqlite3)
  • MCP: @modelcontextprotocol/sdk

Features

  • Multiple persistent chat sessions with SQLite storage
  • Real-time streaming responses from AI providers
  • Model Context Protocol (MCP) integration for tool execution
  • Chat branching and turn-based history
  • Multimodal messages (images, documents)
  • Debug panels with API request/response inspection
  • Settings profiles for different AI providers
  • Anthropic thinking mode and Google Gemini thinking support

Quick Start

Prerequisites

  • Node.js
  • An AI API (Ollama, OpenAI, or any OpenAI-compatible endpoint)
  • Linux: FUSE (required to run the AppImage)

Installation

npm install
npm run dev

Build

npm install
npx electron-builder install-app-deps
npm run build

Output goes to dist-electron/.

Architecture

src/js/
├── app/          # Core logic (message sending, API calls, settings)
├── chat/         # Chat modes and history management
├── tools/        # MCP server management and tool handling
├── render/       # Message rendering and streaming
└── ui/           # Settings, debug panels, context menus

backend/
├── server.js           # Express server (runs inside Electron)
├── config/             # Database initialization
├── routes/             # API endpoints
├── services/           # Chat, MCP, settings, tool events
└── utils/              # Logging

API Compatibility

Works with any OpenAI-compatible API:

  • Ollama - Local AI models
  • OpenAI API - GPT models
  • Anthropic - Claude (via proxy or direct)
  • Google Gemini - Gemini models
  • LM Studio / vLLM - Local inference servers

CI/CD

Pushes to main trigger automated builds via GitHub Actions:

  • Build job: Runs on Windows, Linux, and macOS in parallel. Produces distributable packages (.zip, .AppImage, .dmg).
  • Release job: Triggers only when the version in package.json differs from the latest git tag. Creates a draft GitHub Release with all platform builds attached.

Development

  • Edit files in src/js/ or backend/
  • Reload the app (Ctrl+R) for frontend changes
  • Restart the app for backend changes

Search skills and MCP servers

Search across 31,816 skills and MCPs