twitch-mcp

Twitch MCP Server

AI-powered tools for Twitch streamers, exposed via the Model Context Protocol (MCP). Connect your coding/chat assistants (Gemini CLI, Qwen Coder, Claude Code, etc.) to your Twitch chat for moderation, stream management, and engagement.

Quick Start (No Cloning Required)

Prerequisites

• Node.js 14+ and Java 11+ available on your system PATH
• A Twitch account with appropriate API credentials

Setup Configuration

Create a config file with your Twitch credentials (use bare access token, no "oauth:" prefix):

Windows: C:/Users/<you>/AppData/Roaming/twitch-mcp/config.json
macOS: ~/Library/Application Support/twitch-mcp/config.json
Linux: ~/.config/twitch-mcp/config.json

Example config.json:

Recommended AI CLI Tools

🥇 Primary Recommendation: Gemini CLI

• Excellent MCP support with reliable connections
• Free tier with generous limits
• Great for day-to-day Twitch moderation and chat management

🥈 Secondary Recommendation: Qwen Coder CLI

• Strong coding-focused AI with good MCP integration
• Particularly good for stream development and technical discussions
• Free and open-source

⚡ For Power Users: Claude Code

• Most advanced reasoning capabilities
• Best for complex moderation decisions and nuanced chat analysis
• Note: Limited credits - save for heavy lifting tasks

Configuration Examples

Gemini CLI Configuration

Add to your Gemini settings:

Qwen Coder Configuration

Add to your Qwen settings:

Features

• Chat Management: Send and read messages, recent chat log, chat analysis
• Moderation: Timeout/ban users (by username or descriptor keywords)
• Stream Management: Update title/category, create clips
• Interactive Tools: Create polls and predictions for viewer engagement

Installation Options

Option 1: NPX (Recommended)

No installation required - your AI tool will automatically fetch the latest version:

Option 2: Global Install

Option 3: Local Development

See README-developers.md for building from source.

Usage Notes

• Use a bare access token in auth field (no oauth: prefix needed)
• Only one AI client should connect at a time to avoid conflicts
• Ensure Node.js and Java are installed on the same system as your AI CLI
• The server will automatically handle IRC formatting and API authentication

Troubleshooting

Authentication Issues:

• Confirm auth is a bare token (no oauth: prefix)
• Verify clientId and scopes match your generated token
• Ensure broadcasterId matches your channel ID

Connection Problems:

• Check that both Node.js and Java are in your system PATH
• Try running java -version and node -version to verify installation
• For Windows users: Make sure you're not running from WSL if your AI CLI is on Windows

NPX Prompts in Headless Environments:

• The configurations above use -y flag for non-interactive execution
• If you encounter prompts, add --yes to the npx command

Support

• For general usage questions, check the troubleshooting section above
• For development and contributions, see README-developers.md
• For Claude-specific integration help, see CLAUDE.md