Shell GPT (sgpt) is a command-line productivity tool that puts large language models inside your terminal: describe what you want in plain words and it writes the shell command, explains the error, drafts the code snippet or answers the question, right where you already work. It has earned about 12,300 stars on GitHub and remains actively maintained in 2026.
Shell GPT wraps models such as GPT-5 in an interface designed for terminal habits rather than chat windows. Ask for a command and it generates one you can review, edit and execute in place through shell integration for bash, zsh and PowerShell; describe an error and it suggests the fix; switch to chat mode and it holds a conversation with memory of previous turns, or to REPL mode for an interactive session. Custom roles let you define reusable personas โ a Linux sysadmin, a code generator, a commit-message writer โ and invoke exactly the behavior a task needs. It is not locked to one vendor: OpenAI models work out of the box, and local models served through tools such as Ollama or LiteLLM-compatible endpoints can power it instead, keeping prompts on your own machine. For developers who live in the shell, it removes the context switch of opening a browser every time they forget a flag or need a one-liner.
Shell GPT is free and open source under the MIT license. Your only cost is whatever model you connect โ an OpenAI API key billed by usage, or a local model running on your own hardware at no cost.
Shell GPT is for developers, sysadmins and DevOps engineers who spend their day in a terminal and want LLM help โ commands, code, explanations โ without leaving the shell or memorizing obscure syntax.