AI Agents
OSWorld-V2
Benchmark computer-use agents on long, real-world desktop tasks.
★ 353 starsApache-2.0
View guide →OSWorld — benchmark multimodal agents on open-ended tasks in real computer environments.
OSWorld is an open-source ai agents project by xlang-ai (xlang-ai/OSWorld on GitHub). It has 3,184 GitHub stars and 540 forks, is released under the Apache-2.0 license, and was last updated on GitHub on 14 September 2026.
Decide the first concrete outcome you want from it, so you can pick the right example, template or mode instead of exploring everything at once.
Clone the repository (or install the published package) and skim its README quick start for the requirements for your platform.
Run the documented install step and set any required configuration — such as provider keys, data sources or local services — before the first run.
Execute the smallest documented example end to end, confirm the output is what you expected, then adapt it to your real data or workflow.
Agents act in genuine OS desktops, not simulations or screenshots-only sandboxes.
Run on VMware/VirtualBox locally, Docker with KVM on servers, or Modal, Daytona and AWS sandboxes.
Reference agents to compare your computer-use agent against.
Run the official evaluation locally or via the public evaluation path, with per-domain scores.
Inspect and run individual tasks by hand to debug agent behaviour.
A large suite of open-ended tasks across desktop applications and workflows.
Get the source so you can inspect, configure and run it locally.
git clone https://github.com/xlang-ai/OSWorld.git
cd OSWorldClone, optionally create the conda environment, and install requirements.
git clone https://github.com/xlang-ai/OSWorld
cd OSWorld
pip install -r requirements.txt # or: pip install desktop-envLaunch an experiment with default settings, customising provider and VM path as needed.
python quickstart.py # see README Quick Start for provider options, e.g. --provider_name modal --headless trueIt is maintained by xlang-ai on GitHub (xlang-ai/OSWorld), where the source, issues and releases live.
Yes — it is free and open source under Apache-2.0, which permits use, modification and commercial redistribution, with licence/notice retention.
OSWorld itself is free research code; evaluating an agent needs that agent's model/API access, and VM or cloud sandboxes may incur provider costs.
Compared with productivity computer-use products such as desktop agent apps, OSWorld's distinguishing focus is: oSWorld — benchmark multimodal agents on open-ended tasks in real computer environments. Pick it when that focus matches your need; the alternative may suit you better if its broader or different scope is the priority.
Clone https://github.com/xlang-ai/OSWorld (or install its published package), follow the README quick start to install and configure it, then run the smallest documented example — the Run section on this page mirrors those steps.
Benchmark computer-use agents on long, real-world desktop tasks.
An open benchmark for browser AI agents on everyday tasks.
WorkArena: ServiceNow's benchmark for how well web agents handle everyday knowledge-work tasks.
Tutti — a real-time shared workspace where people and multiple AI agents build in tune, sharing context, files, apps and tasks
ByteDance's open-source multimodal AI agent stack: GUI agents that see and operate real desktops
A benchmark testing whether agents can solve long-horizon coding tasks.
The most-starred other AI Agents projects in the LibreAI catalogue — the realistic shortlist if OSWorld is not the right fit.
A personal AI that really does things — on any OS, with your data staying yours.
An agentic skills framework & software development methodology that works.
Agent harness optimization system with skills, memory and security for Claude Code, Codex and more
The agent that grows with you — a Python framework for adaptive AI agents.
DeepSeek's plugin-based agent harness where everything is a plugin.
Free and open — grab the source, star the repo, and start creating.