ask

AI-powered CLI assistant — modern Unix meets AI

ask is a fast, ergonomic command-line tool that bridges traditional Unix workflows with AI capabilities. Get instant command suggestions, query system information naturally, and tap into AI for complex questions — all from your terminal.

Features

• Multi-provider support — Works with Anthropic, OpenAI, Gemini, Ollama, Perplexity, Groq, Mistral, Cohere, Together, or custom endpoints
• Command suggestions — Get the right command without leaving the terminal
• Natural language system queries — Ask "what is using port 8080" instead of remembering lsof flags
• Streaming AI responses — Real-time output, no waiting for complete responses
• Zero friction — No quotes needed for queries, works with pipes, respects NO_COLOR
• Fast — Native Rust binary, ~3MB, starts instantly
• Secure — Security warnings for untrusted providers, no accidental data leaks

Installation

From crates.io (recommended)

cargo install ask-cmd

Pre-built binaries

Download from GitHub Releases.

From source

git clone https://github.com/askterminal-dev/ask.git
cd ask
cargo install --path .

Quick Start

# Set your API key (Anthropic is the default provider)
export ANTHROPIC_API_KEY="sk-ant-..."

# Or save it to config (secure input, not saved in shell history)
ask config api_key

# Start asking
ask how do I find large files
ask what is using port 3000
ask "explain the difference between threads and processes"

Using Other Providers

# Use OpenAI
ask config provider=openai
export OPENAI_API_KEY="sk-..."

# Use local Ollama (no API key needed)
ask config provider=ollama

# Use Groq for fast inference
ask config provider=groq
export GROQ_API_KEY="gsk_..."

Usage

Command Suggestions

Ask how to do something and get the command:

$ ask how do I compress a folder
tar -czvf archive.tar.gz folder/
zip -r archive.zip folder/

$ ask how to find files by name
find /path -name "pattern"
locate filename

$ ask how do I see disk usage by directory
du -sh */
du -h --max-depth=1

System Information

Query system resources directly:

ask system disk          # Disk usage (df -h)
ask system memory        # Memory stats
ask system cpu           # CPU info
ask system ports         # Network ports
ask system uptime        # System uptime
ask system os            # OS version

Natural Language System Queries

Ask about your system in plain English:

$ ask what is using port 8080
COMMAND   PID   USER   FD   TYPE   DEVICE   NODE NAME
node      1234  user   23u  IPv4   0x...    TCP *:8080 (LISTEN)

$ ask what is using memory
# Shows top memory consumers

$ ask what is using cpu
# Shows top CPU consumers

Command Explanation

Get help for any command:

ask explain grep         # Shows grep --help or man page
ask explain tar
ask describe rsync

AI Queries

For anything else, ask Claude:

ask "what is the capital of France"
ask "write a bash one-liner to count lines in all .py files"
ask "explain the difference between TCP and UDP"

Responses stream in real-time.

Interactive Prompts

For use in scripts:

if ask prompt "Deploy to production?"; then
    ./deploy.sh
fi

Returns exit code 0 for yes, 1 for no.

Input Methods

ask how do I list files     # Direct (quotes optional)
ask "complex query here"    # Quoted
echo "query" | ask          # Piped
ask -i                      # Interactive mode

Configuration

Config file location: ~/.config/ask/config.json

ask config show                           # Show current config
ask config path                           # Show config file path
ask config api_key                        # Set API key (secure input)
ask config provider=openai                # Switch provider
ask config model=gpt-4o                   # Set model
ask config max_tokens=2048                # Set max tokens

Supported Providers

Environment Variables

Custom Providers

You can use custom API endpoints, but ask will show a security warning:

$ ask config api_url=https://my-custom-llm.example.com/v1/chat

SECURITY WARNING

You are configuring a custom api_url that is not in the trusted allowlist:
  api_url = https://my-custom-llm.example.com/v1/chat

This configuration will send your queries and API key to this destination.
Only proceed if you trust this api_url completely.

Type 'yes' to confirm:

This prevents accidental data leaks to untrusted endpoints.

Why ask?

ask keeps you in the terminal and in flow.

Requirements

• macOS or Linux
• Rust 1.70+ (for building from source)
• API key from a supported provider, or local Ollama (only needed for AI queries — system info, command suggestions, and explanations work without any API key)

Contributing

Contributions are welcome! Please see CONTRIBUTING.md for guidelines.

This project follows the Contributor Covenant Code of Conduct.

License

MIT

Built with Rust. Inspired by the wave of modern Unix tools like ripgrep, fd, bat, and eza.