gptme
A personal AI agent for the terminal with shell, Python, web and local models, one of the earliest agent CLIs
Medium risk
We rate an entry medium when the tool runs code, makes network calls or reads project files. Check what exactly it does before installing.
Why this level
- Gives the agent access to a shell and Python REPL on your machine
- Autonomous mode can run commands without constant confirmation
Install
In your terminal, with SkillFoxx CLI
npx skillfoxx add cli/gptmeDetects the agents on your machine, checks the risk and pins the version.
Other ways to install
Install the tool
pipx install gptmeWithout pipx, pip install gptme works too.
Install: pipx install gptme. Run the agent with gptme; for the TUI use pipx install 'gptme[tui]' and the gptme-tui command. As an MCP server: gptme-mcp-server --tools shell,ipython,save,read.
Other ways from the author
pipx install gptmeMCP support is included by default.
This is third-party code. Review the repository files before installing.
What it does
gptme runs an agent right in your terminal, over ssh, in tmux or in CI, giving it shell, Python, web and vision access. It is provider-agnostic: works with Anthropic, OpenAI, Google, xAI, DeepSeek, OpenRouter, or fully locally via llama.cpp. It acts as both an MCP client, loading external servers, and an MCP server that exposes its own shell and Python REPL to Claude Desktop, Cursor and other clients, plus ACP support for editor integration. It has a plugin, skill and lesson system for contextual guidance, a desktop app, and an autonomous agent mode.
Who it is for. For developers who want a vendor-independent terminal agent, including fully local models with no cloud.
Good fit when
- You need a terminal agent that works over ssh or in a headless CI with no GUI
- You want to freely switch model providers, including a fully local one via llama.cpp
- You want to expose your agent's shell and Python REPL as an MCP server for Claude Desktop or Cursor
Not a fit when
- You only need a graphical IDE experience with no terminal
- A ready vendor agent like Claude Code with no provider switching is enough
Example request
Start the local dev server, find the cause of the 500 error in the logs and propose a fixLimitations
Cloud models need your own key for the chosen provider. Autonomous agent mode and background jobs need careful permission setup, otherwise the agent can run unwanted shell commands.
How to disable. Remove the package: pipx uninstall gptme, and remove gptme-mcp-server from your MCP client config if it was connected.
Security check
- Gives the agent access to a shell and Python REPL on your machine
- Autonomous mode can run commands without constant confirmation
README in short
The README presents gptme as a personal agent that runs anywhere a terminal runs: laptop, ssh, tmux, headless servers, CI. It lists tools (shell, Python, web, vision), a plugin and skill system, two-way MCP integration plus ACP, a desktop app for Linux, macOS and Windows, and an autonomous agent mode with background jobs. The project is free and open source, installs via pipx, and has been actively developed since 2023.
FAQ
Can it run fully offline?
Yes, gptme supports fully local models via llama.cpp with no data sent externally.
How is it different from Claude Code or Codex?
gptme is not tied to a single model vendor and was one of the earliest agent CLIs, in active development since spring 2023.
Related
Open-source personal AI assistant on your own machine: answers in Telegram, Slack, Discord and WhatsApp, extended with skills and plugins
A self-improving agent from Nous Research with a TUI, messaging gateway, cron jobs and skills it writes itself
An open source coding agent for the terminal and desktop with build and plan modes
Open prompt library with a Claude Code plugin, MCP server and CLI: search, fetch and improve prompts and skills from an agent