YAICLI

Install YAICLI via pip, pipx, or uv. Use the [all] extra to include all optional provider dependencies, or install specific provider extras like anthropic, gemini, or ollama.

pip install yaicli
# or with all providers:
pip install 'yaicli[all]'
# or with specific providers:
pip install 'yaicli[anthropic,gemini,ollama]'

Run the 'ai' command once to create the default configuration file at ~/.config/yaicli/config.ini. You can view the full template with ai --template.

ai --template

Edit ~/.config/yaicli/config.ini to set your provider and API key. Key config options include PROVIDER (e.g., openai, claude, gemini), API_KEY, MODEL, and BASE_URL. Alternatively, set environment variables like YAI_PROVIDER, YAI_API_KEY, and YAI_MODEL.

# Example ~/.config/yaicli/config.ini entries:
# PROVIDER = openai
# API_KEY = sk-...
# MODEL = gpt-4o

# Or via environment variables:
export YAI_PROVIDER=openai
export YAI_API_KEY=sk-your-key-here
export YAI_MODEL=gpt-4o

Run a one-shot query to verify the installation. Use 'ai -e' for shell command generation mode, or 'ai -c' to enter interactive chat mode.

ai "What is the current directory structure?"
# Execute mode (generates shell commands):
ai -e "list all Python files modified in the last 7 days"
# Interactive chat mode:
ai -c

Add YAICLI to your MCP client's configuration file. YAICLI exposes itself as an MCP server so other agents can call it as a tool. Use the 'ai mcp' subcommand to start the MCP server.

Update your claude_desktop_config.json or equivalent MCP client config to register YAICLI. The server is launched via the 'ai' command with the 'mcp' subcommand.

{
  "mcpServers": {
    "yaicli": {
      "command": "ai",
      "args": ["mcp"],
      "env": {
        "YAI_PROVIDER": "openai",
        "YAI_API_KEY": "sk-your-key-here",
        "YAI_MODEL": "gpt-4o"
      }
    }
  }
}