What is shellgpt?

sgpt is a command-line AI assistant that lives in your terminal. You type a question or instruction in plain English, and it sends that request, along with context about your current directory, operating system, and shell, to an AI provider (like OpenAI) and returns the answer right in your terminal. It works on Linux and macOS with common shells like bash, zsh, and fish. The standout feature is something called "SSH Projection." If you SSH into a remote server, you can bring your local AI assistant along with you, without installing anything on the remote host, without exposing your API key there, and without needing the remote server to have internet access to the AI provider. Your local machine acts as a relay, so the remote server sends requests back through the tunnel and your local machine calls the provider. You can even chain this across multiple hops, like going from your laptop to server A, then from server A to server B, and the AI follows you the whole way. This is built for developers, sysadmins, or anyone who spends a lot of time in the terminal and wants quick help with commands, code analysis, or system tasks. For example, you could ask it to summarize a git diff, analyze what's happening on a remote server you've just SSH'd into, or explain what a flag like -n means in grep. The conversation history persists within a session, so you can ask follow-up questions with sgpt -c. The project is careful about security and resource limits. API keys never leave your local machine, conversation history stays local with strict file permissions, and there are hard caps on input and output sizes to prevent runaway requests. The codebase is written in Rust and includes a robust test suite, but real multi-host testing still requires manual setup.