Fix: command not found

bash: kubectl: command not found
zsh: command not found: docker
sh: 1: node: not found
/bin/sh: python: not found

The shell looked through every directory listed in your PATH variable for an executable with that name and found none. Either the program is not installed, it is installed somewhere PATH does not point, or you typed a name that differs from the real one.

Cause 1: the program is not installed

type kubectl                # "not found" confirms it
apt list --installed 2>/dev/null | grep -i kubectl     # Debian, Ubuntu
rpm -qa | grep -i kubectl                               # RHEL, Fedora
brew list | grep -i kubectl                             # macOS Homebrew

Install it with your package manager and try again. Sometimes the package name is not the command name: pip comes from python3-pip, ifconfig from net-tools, ss and nsenter from iproute2 and util-linux, mkpasswd from whois. On Debian and Ubuntu, apt-file search bin/ifconfig (after apt install apt-file && apt-file update) tells you which package provides a file; on Fedora use dnf provides '*/bin/ifconfig'.

Cause 2: installed, but not on PATH

Tools installed by language package managers land in directories the default PATH does not include: ~/.local/bin (pip), ~/go/bin, ~/.cargo/bin, ~/.npm-global/bin, /usr/local/go/bin, /opt/homebrew/bin on Apple Silicon Macs, or a tool-specific folder.

echo "$PATH" | tr ':' '\n'          # what is searched, in order
find / -name kubectl -type f 2>/dev/null   # where it actually is

# Run it once by full path
~/.local/bin/mytool --version

# Add the directory for this session
export PATH="$HOME/.local/bin:$PATH"

# Make it permanent: add that line to ~/.bashrc (bash) or ~/.zshrc (zsh), then
source ~/.bashrc

A related case: the file exists and is on PATH but is not executable (chmod +x) or has Windows line endings in its first line. Those produce Permission denied or bad interpreter instead, each covered in its own Fix article.

Cause 3: a typo or a different name

Commands are case sensitive and often not what the project is called: python may only exist as python3, Docker Compose v2 is docker compose (a space), nvim not neovim, rg for ripgrep, fd or fdfind on Debian. Tab completion is the quickest check: type the first letters and press Tab twice.

Cause 4: a shell that has not noticed a change

Bash caches command locations. If you installed, moved or deleted a binary in a terminal that is already open, run hash -r or open a new terminal. In cron, CI runners and sudo, PATH is different from your interactive shell: cron has a minimal PATH, and sudo resets PATH through secure_path. Use full paths in those contexts, or set PATH=... at the top of the crontab.

hash -r
sudo env | grep ^PATH=          # what sudo sees
sudo /usr/local/bin/mytool       # full path works regardless

Check that it is fixed

command -v kubectl      # prints the path and exits 0
kubectl version --client

When it is not what you think

• ./script.sh: line 3: foo: command not found inside a script points at a command used by the script, not at the script itself. Often it is a function or alias that only exists in your interactive shell, or $'\r': command not found, which means CRLF line endings.
• sudo: mytool: command not found while mytool works for you: sudo uses its own PATH. Use the full path or sudo -E env "PATH=$PATH" mytool if your policy allows it.
• A command that worked yesterday and is gone today usually means a version manager (nvm, pyenv, rbenv, asdf) is no longer initialized in your shell startup file, or a Python virtual environment is not activated.
• Permission denied is a different error: the file was found but cannot be executed.

Official documentation

• Bash manual: Command Search and Execution
• Bash manual: PATH variable