shellchat

ShellChat

ShellChat is a Rust CLI tool that brings AI to your terminal, enabling natural language command generation and interactive chat with various LLM providers. It supports MCP Servers and multiple AI providers: OpenAI, DeepSeek, Anthropic, Gemini and OpenRouter. Inspired by ShellGPT, it's both a practical tool and a Rust learning project.

Example Usage

Chat with AI in the terminal:

schat "What is the capital of France?"

Convert natural language to shell commands:

schat -s "List all files in the current directory"

Continuous chat mode:

schat --chat

Enter an interactive chat session where you can converse with the AI model continuously. This mode supports:

• Interactive Conversation: Engage in multi-turn dialogues with the AI.
• Slash Commands: Use special commands to manage your chat session:
  • /help: Display a list of all available commands.
  • /clear: Clear the current conversation history.
  • /model <name>: Show or change the active LLM model.
  • /save <filename>: Save the current conversation history to a file.
  • /load <filename>: Load a conversation history from a file.
  • /list: List all saved conversation history files.
  • /delete <filename>: Delete a specific conversation history file.
  • /display <mode>: Control the visibility of tool interactions (modes: verbose, minimal, hidden, help).
• Tool Calling with MCP Servers: When configured, the AI can automatically use tools provided by MCP Servers to perform actions like searching the web or accessing external APIs.

Generate a commit message for git changes:

git diff | schat "Generate a commit message for the changes"

Summary a text file:

schat "Summarize the content of this file" < file.txt

Installation

Quick Install from GitHub Releases

Linux Installation:

# One-line install using curl
curl -sSL https://raw.githubusercontent.com/zwldarren/shellchat/main/scripts/install.sh | bash -s -- install

# Uninstall
curl -sSL https://raw.githubusercontent.com/zwldarren/shellchat/main/scripts/install.sh | bash -s -- uninstall

Windows Installation:

# One-line install using irm
irm https://raw.githubusercontent.com/zwldarren/shellchat/main/scripts/install.ps1 -OutFile $env:TEMP\install.ps1; & $env:TEMP\install.ps1 install

# Uninstall
irm https://raw.githubusercontent.com/zwldarren/shellchat/main/scripts/install.ps1 -OutFile $env:TEMP\install.ps1; & $env:TEMP\install.ps1 uninstall

Configuration

The configuration file is located at ~/.schat/config.yaml.

Configuration Options

Global Settings

# Currently active LLM provider (must match a provider key below)
active_provider: openai

# Whether to auto-confirm shell command execution (true/false)
auto_confirm: false

LLM Providers Configuration

Configure one or more LLM providers. All providers support these common parameters:

• api_key: Your API key for the provider
• base_url: API endpoint URL
• model: Model name to use

providers:
  # OpenAI compatible API
  openai:
    api_key: your_openai_api_key_here
    base_url: https://api.openai.com/v1
    model: gpt-4.1-mini  # Example: gpt-4, gpt-3.5-turbo
  
  # OpenRouter API (supports multiple providers)
  openrouter:
    api_key: your_openrouter_api_key_here
    model: google/gemini-2.0-flash-001
  
  # DeepSeek API
  deepseek:
    api_key: your_deepseek_api_key_here
    model: deepseek-chat
  
  # Google Gemini API
  gemini:
    api_key: your_gemini_api_key_here
    model: gemini-2.0-flash
  
  # Anthropic Claude API
  anthropic:
    api_key: your_anthropic_api_key_here
    model: claude-sonnet-4-20250514

MCP Servers Configuration

Model Context Protocol (MCP) servers extend functionality with additional tools.

mcp_servers:
  # Local stdio-based server
  - name: everything
    enabled: false  # Set to true to enable
    type: stdio     # Local process
    command: npx    # Command to run
    args: 
      - "-y"
      - "@modelcontextprotocol/server-everything"
  
  # Brave Search API server
  - name: brave-search
    type: stdio
    command: npx
    args: ["-y", "@modelcontextprotocol/server-brave-search"]
    envs:
      BRAVE_API_KEY: your_brave_api_key_here  # Required API key
  
  # HTTP-based server
  - name: StreamableHttp-server
    type: streamable-http
    url: https://mcp.example.com/mcp
  
  # SSE-based server
  - name: sse-server
    type: sse
    url: https://mcp.example.com/sse

Supported Providers

Get API keys from these providers:

• OpenAI
• DeepSeek
• Anthropic
• Gemini
• OpenRouter