Chat MCP CLI

A beautiful command-line chat interface that connects to MCP (Model Context Protocol) servers and supports multiple AI providers. Chat with LLMs while leveraging the power of MCP tools directly from your terminal.

Features

• 🤖 Multiple AI Providers: Support for OpenAI, Anthropic, Google, and OpenRouter
• 🔌 MCP Server Integration: Connect to any MCP-compatible server (stdio, HTTP, SSE)
• 🛠️ Tool Calling: Automatically invoke MCP tools during conversations
• 🎨 Beautiful UI: Terminal UI built with React and Ink
• 💬 Interactive Chat: Real-time streaming responses with tool execution display
• ⚙️ Flexible Configuration: JSON-based configuration for both AI providers and MCP servers

Installation

# Clone the repository
git clone https://github.com/abhignakumar/chat-mcp-cli
cd chat-mcp-cli

# Install dependencies
npm install

# Build the project
npm run build

# Run the application
npm start

Configuration

The application requires two configuration files in your working directory:

1. AI Provider Configuration (config.json)

Configure your AI provider and model:

{
	"apiProvider": "openai",
	"model": "gpt-5.2",
	"apiKey": "your-api-key"
}

Supported Providers:

• openai - OpenAI models (GPT-5.2, GPT-5.2-pro, etc.)
• anthropic - Anthropic models (Claude Opus 4.5, etc.)
• google - Google Gemini models
• openrouter - OpenRouter models

See example-config.json for a template.

2. MCP Server Configuration (mcp.json)

Configure your MCP servers:

{
	"servers": {
		"github": {
			"type": "http",
			"url": "https://api.githubcopilot.com/mcp/",
			"headers": {
				"Authorization": "Bearer your-api-key"
			},
			"disabled": false
		},
		"local-server": {
			"type": "stdio",
			"command": "node",
			"args": ["server.js"],
			"env": {
				"API_KEY": "value"
			},
			"disabled": false
		}
	}
}

Server Types:

• http - HTTP-based MCP server
• sse - Server-Sent Events MCP server
• stdio - Local stdio-based MCP server

Options:

• disabled - Set to true to disable a server
• disabledTools - Array of tool names to disable

See example-mcp.json for a template.

Usage

Run the CLI

# From the project directory (with config.json and mcp.json in current directory)
npm start

# Or run the file directly
node dist/cli.js

Commands

• Type your message and press Enter to chat
• Type /exit or /quit to exit the application

Display

The CLI features:

• Welcome Screen: Shows connected MCP servers, current model, and working directory
• Chat History: Displays conversation with user messages and AI responses
• Tool Execution: Shows when tools are called and their results
• Input Box: Interactive text input at the bottom
• Status Indicator: Shows when the AI is processing or streaming

Architecture

• src/cli.tsx - Entry point that renders the React app
• src/app.tsx - Main application component with chat logic and MCP integration
• src/mcp-client.ts - MCP client wrapper for connecting to servers
• src/utils.ts - Utility functions for configuration and AI model setup
• src/types.ts - TypeScript types and Zod validation schemas
• src/components/ - React components for the terminal UI

Dependencies

• Ink - React for interactive command-line apps
• @ai-sdk - AI SDK for multiple providers
• @modelcontextprotocol/sdk - MCP Client SDK
• Zod - Schema validation