ask

simple bash script that helps you with terminal commands with the power of chatgpt

Project maintained by zmsp Hosted on GitHub Pages — Theme by mattgraham

ask — AI terminal assistant

Turn plain-English descriptions into shell commands. Powered by OpenAI, Google Gemini, or local Ollama. Zero external dependencies.

Demo

Features

Command generation — Get shell commands tailored to your OS, architecture, and shell.
Interactive execution — [y]es run, [e]dit in readline before running, [c]opy to clipboard, or [n]o.
ask fix — Diagnose and correct the last failed command or piped error stream.
ask cheat <tool> — Instant 5-recipe cheatsheet for tricky utilities (tar, ffmpeg, jq).
ask !! — Explain the last shell command from history.
Git workflow suite:
- ask commit — Generate clean commit messages from git status & diff.
- ask review — AI code review on uncommitted or branch changes.
- ask branch <task> — Semantic kebab-case branch name generator + checkout.
Local AI support — Run completely offline and free via Ollama (no API key required).
Token & query tracker — ask --stats tracks usage across models and providers.
Inline shell keybinds — ask --init provides Ctrl-X Ctrl-A inline completion for Bash, Zsh, and PowerShell.
--raw scripting flag — Pure command string output for aliases, scripts, and shell integrations.
Dangerous command guard — Warns and requires explicit confirmation for risky patterns (rm -rf, sudo).
0 external dependencies — Written in pure Bash and PowerShell; uses only standard OS utilities.

Quick Install

macOS (Homebrew — Recommended)

brew tap zmsp/ask
brew install ask

macOS / Linux (Installer Script)

curl -fsSL https://zmsp.github.io/ask/install.sh | bash

Or manually:

curl -fsSL https://raw.githubusercontent.com/zmsp/ask/main/ask.sh -o /usr/local/bin/ask
chmod +x /usr/local/bin/ask
ask --setup

Windows (PowerShell)

irm https://zmsp.github.io/ask/install.ps1 | iex

Or manually:

irm https://raw.githubusercontent.com/zmsp/ask/main/ask.ps1 -OutFile "$env:USERPROFILE\bin\ask.ps1"
# Add $env:USERPROFILE\bin to your PATH, then:
ask --setup

Uninstall

# Homebrew
brew uninstall ask

# macOS / Linux installer
curl -fsSL https://zmsp.github.io/ask/install.sh | bash -s -- --uninstall

# Windows (PowerShell)
& ([scriptblock]::Create((irm https://zmsp.github.io/ask/install.ps1))) -Uninstall

Platform	Script	Installer
macOS (Brew)	`Formula/ask.rb`	`brew install zmsp/ask/ask`
macOS / Linux	`ask.sh` (bash)	`install.sh`
Windows	`ask.ps1` (PowerShell 5.1+)	`install.ps1`

Requirements

Tool	Install
`bash` 3.2+ or `pwsh`	Pre-installed on macOS/Linux/Windows
`curl`	`brew install curl` / `apt install curl`
`jq`	`brew install jq` / `apt install jq` (Bash script only)
AI provider	OpenAI, Gemini, or Ollama (local)

Usage

Generate & run a shell command

ask "list all .log files modified in the last 7 days"
ask "restart nginx if it's not running"
ask "find the 5 largest files in my home directory"

Suggested:
find ~ -type f -mtime -7 -name "*.log"

Run? [y]es / [e]dit / [c]opy / [n]o: e
Edit: find ~ -type f -mtime -14 -name "*.log"

y runs the command immediately.
e lets you edit the command inline in terminal readline before executing.
c copies it directly to your clipboard (pbcopy, wl-copy, xclip, or clip.exe).
n cancels.

Diagnose & fix errors (`ask fix`)

# Fix last failed command from shell history
docker run -p 80:80 myapp
# Error: port 80 is already allocated
ask fix

# Or pipe stderr directly
cargo build 2>&1 | ask fix
git push 2>&1    | ask fix

CLI Cheatsheet (`ask cheat <tool>`)

ask cheat tar
ask cheat ffmpeg
ask cheat jq

Explain the last command (`ask !!`)

tar -czvf backup.tar.gz --exclude="*.log" ./data
ask !!
# → Explains flags, paths, compression behavior, and risks

Code review & Git helpers

# Concise senior review of current uncommitted or branch diff
ask review

# Generate semantic branch name and switch to it
ask branch "add user password reset with email token"
# → Suggested: feat/user-password-reset
# Create and switch to this branch? [y]es / [c]opy / [n]o: y

# AI-powered commit message generator
ask commit

Free-form AI question

ask -q "What does HEAD~3 mean in git?"
ask -e "Explain the difference between CMD and ENTRYPOINT in Docker"

Pipe content as context

cat error.log  | ask "why is this failing?"
git diff       | ask "summarise these changes"
cat config.yml | ask "is anything misconfigured here?"

Track token usage & stats

ask --stats

ask · Usage Statistics
Tracked in: ~/.ask_stats

Total queries:      14
Prompt tokens:      2,350
Completion tokens:  410
Total tokens:       2,760

Queries by provider/model:
  openai (gpt-4.1-nano)            10
  ollama (llama3.2)                4

Shell Keybinding (Inline AI expansion)

Press Ctrl-X Ctrl-A on any line to replace what you typed with the AI-suggested command.

# Print keybind configuration for your shell:
ask --init bash  # or zsh or pwsh

# Quick install into ~/.zshrc:
eval "$(ask --init zsh)"

Configuration

Configuration is stored in ~/.ask_config (mode 600, user-readable only).

Run the interactive wizard:

ask --setup

Or edit directly:

# ~/.ask_config
provider=openai          # openai | gemini | ollama
model=gpt-4.1-nano       # leave blank for cheapest default
max_tokens=200
ollama_endpoint=http://localhost:11434/v1
openai_api_key=sk-...
gemini_api_key=AIza...

Environment variables

Environment variables take priority over config file values:

export OPENAI_API_KEY="sk-..."
export GEMINI_API_KEY="AIza..."
export VERBOSE=true           # print debug info

Provider & Model Reference

Provider	Default model	Price (input/output per 1M tokens)
`openai`	`gpt-4.1-nano`	$0.10 / $0.40
`gemini`	`gemini-2.5-flash-lite`	$0.10 / $0.40
`ollama`	`llama3.2`	Free (runs locally)

OpenAI models: gpt-4.1-nano · gpt-4.1-mini · gpt-4.1 · gpt-4o · gpt-4o-mini · o3-mini
Gemini models: gemini-2.5-flash-lite · gemini-2.5-flash · gemini-2.5-pro · gemini-3.5-flash-lite · gemini-3.8-flash · gemini-flash-latest (latest) · gemini-flash-lite-latest · gemini-pro-latest
Ollama models: llama3.2 · qwen2.5-coder · codellama · any local model

ask