Open Terminal CLI
Open Terminal gives AI agents a computer of their own with a REST API: shell, files, code, browser and desktop
- 3,559 ★ on the repository
- Last commit
- MIT
Broad access
Why this level
- Executes arbitrary commands and code with an API key
- Outside Docker commands run with your user's permissions
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add cli/open-terminalDetects the agents on your machine, shows the access level and pins the version.
Other ways to install
Access and checks
How we reviewBroad accessThe tool can change data, run commands or handle secrets.
- Executes arbitrary commands and code with an API key
- Outside Docker commands run with your user's permissions
This is third-party code. Review the repository files before installing.
What it does
Open Terminal is a self-hosted service that hands agents a real computer through a REST API. The agent runs commands, searches and changes files, executes code and works with browsers and desktops. It connects to Open WebUI: add an address and key in admin settings, then select the terminal in chat. The easiest way is Docker: the `latest` image carries Python, Node.js, build tools, a browser and document tools, `agents` adds the latest Codex and Claude Code CLIs, `cuda` adds GPU support, and `slim` and `alpine` are minimal. A Python install exists too: `pip install open-terminal`.
Who it is for. Open WebUI administrators and agent developers who need a separate sandbox with a shell, filesystem and browser.
Good fit when
- A chat in Open WebUI must run code and work with the user's files.
- An agent builds a project, runs tests and downloads packages in an isolated environment.
- Codex or Claude Code is launched from chat in a container with your own sign-in.
- An agent needs a browser or desktop to check web pages.
- Files between container runs are kept on a named volume.
Not a fit when
- The API key would go to people or agents you do not trust: it grants command and file access.
- The service runs on a working host without isolation: commands execute with your user's permissions.
- You only need a tool list with no server of your own.
Example request
Clone the repository into the terminal, install dependencies, run the tests and show which ones failed.Limitations
The `OPEN_TERMINAL_API_KEY` grants command and file access: outside Docker, commands run with your user's permissions. If no key is set, one is generated and printed at startup, and in Docker you find it with `docker logs open-terminal`. A Python install needs version 3.11 or newer. Codex and Claude Code inside the image sign in with your own account. The `latest` image is large because it includes a browser and a desktop.
How to disable. Stop and remove the container with `docker stop open-terminal` and `docker rm open-terminal`, and disable the integration in Open WebUI. The `open-terminal` volume with files is deleted separately.
FAQ
How do I run it in Docker?
With `docker run -d --name open-terminal --restart unless-stopped -p 8000:8000 -v open-terminal:/home/user -e OPEN_TERMINAL_API_KEY=your-secret-key ghcr.io/open-webui/open-terminal:latest`. The API opens at `http://localhost:8000`, docs at `/docs`.
How do I connect it to Open WebUI?
In admin settings open Integrations, then Open Terminal, add an address reachable from the Open WebUI server and the key, enable the connection and select the terminal in chat.
How do the images differ?
`latest` is the full set, `agents` adds Codex and Claude Code, `cuda` brings GPU, `slim` and `alpine` are minimal and `openshift` suits restricted policies.
Can I run it without Docker?
Yes, `pip install open-terminal`, then `open-terminal run --host 0.0.0.0 --port 8000`. Python 3.11 or newer is required.
How do I avoid an open port?
Run with `--isolated --ipc /path/api.sock`: desktop clients connect through a private socket and the key is still required.
Discussion
No questions or reviews yet.
Related
CLI toolsA Rust terminal with a multiplexer: agent sessions from several machines in one window, plus a skill for seeing adjacent panes
A Docker sandbox for coding agents: a Debian image with Copilot CLI, Codex and Pi, and SSH, RDP and Docker enabled by flags.
A native macOS terminal built on Ghostty with vertical tabs and notifications when an agent needs your input
An htop-like terminal panel for AI agents: Claude Code, Codex, OpenCode and Devin sessions, CPU, tokens and cost.
CLI that runs commands and AI agents in isolated microVMs that boot in under 125 ms and keeps API keys out of the sandbox itself.
AgentBox runs Claude Code, Codex and Open Code in separate virtual machines on your PC, a self-hosted server or a cloud.
README in short
The README covers Docker and Python installs, a private local socket, connecting to Open WebUI and launching coding agents from chat.