tgpt is a cross-platform command-line tool, written in Go, that lets you chat with AI models directly from your terminal — with no API key required. It connects to freely accessible model providers so you can ask questions, generate code and draft shell commands without accounts, billing or configuration files. The project has roughly 3,300 stars on GitHub and remains actively maintained, with updates landing in 2026.
The experience is deliberately frictionless: install a single binary on Windows, macOS or Linux and start a conversation, or use flags for one-shot prompts. tgpt can read piped stdin, which makes it natural to summarize logs, explain error output or transform text inside shell pipelines. Dedicated modes generate shell commands from a plain-English description (with a chance to review before running), produce code snippets, and even create images from text prompts. An interactive chat mode keeps conversation context, and a provider system lets you choose among several backends and models — including GPT-4-class, Llama and Mixtral models offered through free tiers — or point the tool at your own API key if you prefer. Everything is configured through simple flags and environment variables, keeping the tool scriptable and automation-friendly.
tgpt is completely free and open source under the GPL-3.0 license, and its default providers work without payment. Optional self-supplied API keys use your own provider quota.
tgpt is for terminal-centric developers, sysadmins and curious users who want quick AI answers, command drafting and text processing in the shell without signing up for anything.