terminalGPT

We'd reach for terminalGPT when the alternative is alt-tabbing to a web chat to ask something about the code on screen. The install is one command, the first-run flow is just pasting a key, and after that `tgpt chat` is the whole interface. That's the pitch, and it holds up. The multi-provider angle is what separates it from most terminal chat tools we've tried. Having OpenAI, Anthropic, Groq, and Gemini behind one CLI means you aren't rewriting your habits when you switch models. Practically, that matters most for Groq's fast inference on trivial questions and for keeping a Claude key around when you want a longer explanation. Where it bites: this is a text-in, text-out loop. Everything happens inside the terminal session, so if your workflow depends on dragging a screenshot or a PDF into the conversation, you'll be doing that somewhere else. Teams should also look past it — there's no shared history, no role-based access, and no audit trail, which is fine for personal keys and wrong for regulated work. Compared with heavier terminal AI tools, terminalGPT stays small: TypeScript source, no background daemons, MIT license, 1k stars and 178 forks on GitHub. Compared with shell-integrated assistants that wire AI into your prompt itself, it's more of a standalone chat client — you invoke it, talk, and exit. That's a feature if you want separation, a limitation if you want the model reading your shell context. Our take on picking it: choose terminalGPT if you want provider choice plus local conversation storage and you're comfortable managing your own API keys and spend. Pass if you need multimodal input, collaborative threads, or an assistant that lives inside your editor and understands your project tree. For a solo developer who spends the day in a shell, the