gptme
A free, open-source personal AI agent that runs in any terminal, including SSH sessions, tmux and CI pipelines, with built-in shell, Python, web and vision tools, and a choice of hosted or local models.
gptme is a personal AI agent that runs anywhere a terminal runs, provider-agnostic and local-first, shipping with shell, Python, web browsing, vision and other tools an agent needs out of the box. It works with hosted providers including Anthropic, OpenAI, Google and DeepSeek, an existing ChatGPT or SuperGrok subscription, or fully local inference through Ollama or any OpenAI-compatible server.
It supports plugins, skills and lessons for extensibility, MCP and ACP integrations, autonomous agent runs with guardrails, and a desktop app for Linux, macOS and Windows alongside its command-line interface. One of the earliest agent CLIs, dating to spring 2023, it remains in active development, with recent releases adding cross-harness memory shared with Claude Code and Codex, a terminal UI, and sandboxed Python and shell execution through Docker or Wasmtime.