gptme is a free, open-source personal AI agent that lives in your terminal. It gives a language model real local tools โ a shell, Python, file editing, web browsing and vision โ so it can write code, run commands and carry out multi-step knowledge work on your own machine.
First released in spring 2023, gptme is one of the earliest agent CLIs and is still in very active development. It is provider-agnostic and local-first: it works with Anthropic, OpenAI, Google, DeepSeek and other hosted models, through aggregators such as OpenRouter, with an existing ChatGPT or Grok subscription, or fully locally via Ollama and any OpenAI-compatible server. Beyond the classic CLI it now ships a terminal UI, a web UI with a REST API server, and a desktop app. Its extensibility system layers plugins, skills, contextual lessons and hooks, and a cross-harness memory store lets gptme share persistent memory with other agent tools. It speaks MCP and ACP for tool and harness interoperability, supports autonomous and headless operation (including scheduled agents under systemd), and includes guardrails for unattended runs. gptme is MIT-licensed, has more than 4,400 stars on GitHub, and was under active development in 2026 with frequent releases.
Free. gptme is open source under the MIT license; you only pay for any hosted model APIs you choose to connect.
Developers and power users who want a capable, hackable coding and general-purpose agent that runs anywhere a terminal runs โ laptop, SSH session, tmux or CI โ without giving up control of their models or data.