Open-source multimodal AI agent stack from ByteDance, with a CLI/Web UI (Agent TARS) and a desktop GUI agent app (UI-TARS Desktop).
e2b-dev/desktop
An open source Linux desktop sandbox on E2B that AI models can control through mouse, keyboard and screenshots, with a streamable view.
- Stars
- 1,557
- Forks
- 188
- Open issues
- 11
- Last commit
- yesterday
What is it?
E2B Desktop Sandbox is a virtual Linux desktop environment built on E2B Sandbox, designed for computer use by LLMs. Each sandbox is isolated and can be customized with additional dependencies. Developers create a sandbox from the Python or JavaScript SDK, then launch applications such as Chrome, Firefox or VS Code inside it.
The SDK exposes control primitives for the desktop: mouse clicks, drags, scrolling, typing, key combinations, window queries, screenshots and shell commands. The desktop can be streamed to a browser URL, either as the full screen or a single application window, with optional password protection and view-only mode.
The desktop runs on Linux with the Xfce environment. The sandbox template is open for modification, so teams can build a custom template with extra packages or a different desktop environment.
Key features
- 01Mouse, keyboard and scroll control via Python and JavaScript SDKs.
- 02Screenshot capture and window querying for agent feedback loops.
- 03Streaming of the full desktop or a single application window to a URL.
- 04Password-protected and view-only stream options.
- 05Customizable template built on Linux and Xfce.
Use cases
- Letting an LLM agent operate a GUI application that has no API.
- Running browser-based computer use tasks in an isolated environment.
- Streaming a remote desktop session for demos or human supervision.
- Building custom desktop templates with specific tools preinstalled.
Quick start
- 1
Get an E2B API key and set E2B_API_KEY
- 2
Install the Python SDK
pip install e2b-desktop - 3
Create a desktop sandbox and launch an application
Things to know
- Requires an E2B account and API key to run sandboxes.
- Only one stream can be active at a time.
In collections
Topics
FAQ
- What is desktop?
- An open source Linux desktop sandbox on E2B that AI models can control through mouse, keyboard and screenshots, with a streamable view. E2B Desktop Sandbox is a virtual Linux desktop environment built on E2B Sandbox, designed for computer use by LLMs. Each sandbox is isolated and can be customized with additional dependencies. Developers create a sandbox from the Python or JavaScript SDK, then launch applications such as Chrome, Firefox or VS Code inside it.
- Is desktop free to use? What license does it have?
- desktop is open source under the Apache-2.0 license. Check the license terms before using it in commercial products.
- Is desktop still maintained?
- Actively maintained. The last commit to desktop was on Nov 6, 2026.
- What can I use desktop for?
- Letting an LLM agent operate a GUI application that has no API. Running browser-based computer use tasks in an isolated environment. Streaming a remote desktop session for demos or human supervision. Building custom desktop templates with specific tools preinstalled.
- How do I install desktop?
- Get an E2B API key and set E2B_API_KEY Install the Python SDK: pip install e2b-desktop Create a desktop sandbox and launch an application
- Where can I get desktop?
- The source code and installation instructions are on GitHub: https://github.com/e2b-dev/desktop
Summary generated by AI from the project's README. GitHub stats refresh daily.
Related projects
An open-source computer use agent framework that controls a real GUI with mouse, keyboard and screen to complete tasks in desktop and web apps.
A library and hosted service that gives AI agents isolated Linux-like sandboxes with files, processes, shell, and networking, built on WebAssembly and V8.
A framework for autonomous Android and computer use, letting a planner and finder LLM operate phones and desktops from natural-language tasks.
A streaming server for Moonlight that lets multiple remote clients play games on one shared host, with each session running in Docker or Podman containers.
A runtime interface that lets AI agents run shell commands in sandboxed environments, locally or on cloud platforms like Docker, AWS, and Modal.