ShellGPT — AI Assistant in Your Terminal

What it is

ShellGPT is a command-line tool that generates shell commands, code, and documentation from natural language prompts. You type what you want in plain English, and ShellGPT returns the exact command or script to run. It supports chat mode for multi-turn conversations, a REPL for interactive sessions, and local model backends.

It targets developers, sysadmins, and DevOps engineers who spend significant time looking up CLI syntax, writing one-off scripts, or translating intent into shell commands.

How it saves time or tokens

Instead of searching Stack Overflow or man pages for the right flags, ShellGPT gives you the command immediately. The shell integration means you can execute generated commands without copy-pasting. For repetitive tasks like log parsing, file manipulation, or system diagnostics, ShellGPT turns a 5-minute search into a 5-second prompt. The token estimate for a typical session is around 297 tokens per interaction.

How to use

1. Install ShellGPT:

pip install shell-gpt

1. Set your API key and run a query:

export OPENAI_API_KEY=sk-...
sgpt 'find all Python files modified in the last 7 days'

1. Use shell mode for executable commands:

sgpt --shell 'compress all .log files in /var/log older than 30 days'
# ShellGPT returns: find /var/log -name '*.log' -mtime +30 -exec gzip {} \;
# [E]xecute, [D]escribe, [A]bort?

Example

Using the REPL for multi-turn code generation:

sgpt --repl code
>>> Create a Python function that reads a CSV and returns rows where column 'status' is 'active'

import csv

def get_active_rows(filepath):
    with open(filepath) as f:
        reader = csv.DictReader(f)
        return [row for row in reader if row['status'] == 'active']

>>> Now add error handling for missing files

import csv
from pathlib import Path

def get_active_rows(filepath):
    path = Path(filepath)
    if not path.exists():
        raise FileNotFoundError(f'{filepath} not found')
    with open(path) as f:
        reader = csv.DictReader(f)
        return [row for row in reader if row['status'] == 'active']

Related on TokRepo

• Automation tools — CLI and workflow automation resources
• AI coding tools — AI-powered development assistants

Common pitfalls

• Running generated commands without review can be dangerous. Always use the describe option before executing destructive operations like rm -rf or DROP TABLE.
• ShellGPT defaults to OpenAI models. To use local models, configure the backend URL to point to Ollama or other compatible endpoints.
• Chat history grows quickly in REPL mode. Long sessions accumulate context that increases token usage and cost.