AI CLI Chat

A terminal-based AI chat interface built with React 18, Ink, and Deno that provides interactive conversations with OpenAI's GPT models and includes MCP (Model Context Protocol) tool support.

Features

• Interactive Chat Interface - Real-time conversations with AI
• MCP Tool Integration - File operations, directory listing, and shell commands
• Built with Modern Stack - React 18, Ink, Deno, TypeScript
• Terminal UI - Clean, responsive terminal interface
• Safe Exit - Double Ctrl+C or Escape to prevent accidental exits
• File System Access - Read, write, and list files through AI commands

Prerequisites

• Deno installed
• OpenAI API key

Setup

1. Clone or download this project
2. Set your OpenAI API key as an environment variable:

   export OPENAI_API_KEY="your-api-key-here"
   

Usage

Start the application

deno task start

Development mode (with hot reload)

deno task dev

Controls

• Enter - Send message
• Ctrl+C twice - Exit application
• Escape twice - Exit application

Available AI Tools

The AI assistant has access to these MCP tools:

File Operations

• read_file - Read contents of any file
• write_file - Write content to files
• list_files - List directory contents
• run_command - Execute shell commands

Example Commands

Try asking the AI:

• "List the files in this directory"
• "Read the contents of main.tsx"
• "Create a new file called test.txt with some example content"
• "Run the command 'git status'"

Project Structure

llm-cli/
├── main.tsx          # Main React application
├── tools.ts          # MCP tool definitions and execution
├── deno.jsonc        # Deno configuration and dependencies
└── README.md         # This file

Configuration

deno.jsonc

The project configuration includes:

• React 18 with JSX runtime
• Ink 3.2.0 for terminal UI
• OpenAI SDK for GPT integration
• MCP SDK for tool protocol support

Environment Variables

• OPENAI_API_KEY - Required for OpenAI API access

Dependencies

• react@^18.2.0 - UI framework
• ink@^3.2.0 - Terminal UI components
• openai@^4.47.1 - OpenAI API client
• @modelcontextprotocol/sdk@^0.5.0 - MCP protocol support
• ink-text-input@5 - Terminal input component

Architecture

The application follows a component-based architecture:

1. Main App (main.tsx) - Handles UI state, user input, and AI interactions
2. Tools Module (tools.ts) - Defines and executes MCP tools
3. OpenAI Integration - Implements function calling with tool support

Message Flow

1. User types message and presses Enter
2. Message added to conversation history
3. OpenAI API called with tools and conversation context
4. If AI decides to use tools, they're executed automatically
5. Results displayed in the terminal with tool indicators (🔧)

Development

Adding New Tools

1. Define tool schema in tools.ts
2. Add execution logic to executeTool() function
3. Tools are automatically available to the AI

Modifying UI

The terminal interface uses Ink components. Key components:

• Box - Layout containers
• Text - Styled text output
• TextInput - User input field

Security

• Tools execute with Deno's permission system
• File operations require --allow-read and --allow-write
• Shell commands require appropriate permissions
• API key handled through environment variables

Troubleshooting

Common Issues

"Module not found" errors

• Run deno cache main.tsx to download dependencies

"Permission denied" errors

• Ensure you're running with appropriate Deno permissions:

  deno run --allow-env --allow-run --allow-read --allow-write --allow-net main.tsx
  

API key errors

• Verify your OPENAI_API_KEY environment variable is set
• Check that your API key has sufficient credits

JSX runtime errors

• Ensure you're using the provided deno.jsonc configuration