Topics

#agent-computer-interface #ai-agents #computer-automation #computer-use #computer-use-agent #cua #grounding #gui-agents #in-context-reinforcement-learning #memory #mllm #planning #retrieval-augmented-generation

Language breakdown

No language data available.

▸ Files click to expand

README

Agent S: Use Computer Like a Human

🏆 Agent S3: First to Surpass Human Performance on OSWorld (72.60%)

🌐 [S3 blog] 📄 [S3 Paper] 🎥 [S3 Video]

🌐 [S2 blog] 📄 [S2 Paper (COLM 2025)] 🎥 [S2 Video]

🌐 [S1 blog] 📄 [S1 Paper (ICLR 2025)] 🎥 [S1 Video]

Skip the setup? Try Agent S in Simular Cloud

🥳 Updates

[x] 2025/12/15: Agent S3 is the first to surpass human-level performance on OSWorld with an impressive score of 72.60%!
[x] 2025/10/02: Released Agent S3 and its technical paper, setting a new SOTA of 69.9% on OSWorld (approaching 72% human performance), with strong generalizability on WindowsAgentArena and AndroidWorld! It is also simpler, faster, and more flexible.
[x] 2025/08/01: Agent S2.5 is released (gui-agents v0.2.5): simpler, better, and faster! New SOTA on OSWorld-Verified!
[x] 2025/07/07: The Agent S2 paper is accepted to COLM 2025! See you in Montreal!
[x] 2025/04/27: The Agent S paper won the Best Paper Award 🏆 at ICLR 2025 Agentic AI for Science Workshop!
[x] 2025/04/01: Released the Agent S2 paper with new SOTA results on OSWorld, WindowsAgentArena, and AndroidWorld!
[x] 2025/03/12: Released Agent S2 along with v0.2.0 of gui-agents, the new state-of-the-art for computer use agents (CUA), outperforming OpenAI's CUA/Operator and Anthropic's Claude 3.7 Sonnet Computer-Use!
[x] 2025/01/22: The Agent S paper is accepted to ICLR 2025!
[x] 2025/01/21: Released v0.1.2 of gui-agents library, with support for Linux and Windows!
[x] 2024/12/05: Released v0.1.0 of gui-agents library, allowing you to use Agent-S for Mac, OSWorld, and WindowsAgentArena with ease!
[x] 2024/10/10: Released the Agent S paper and codebase!

💡 Introduction

Welcome to Agent S, an open-source framework designed to enable autonomous interaction with computers through Agent-Computer Interface. Our mission is to build intelligent GUI agents that can learn from past experiences and perform complex tasks autonomously on your computer.

Whether you're interested in AI, automation, or contributing to cutting-edge agent-based systems, we're excited to have you here!

🎯 Current Results

Agent S3 Results

On OSWorld, Agent S3 alone reaches 66% in the 100-step setting, already exceeding the previous state of the art of 63.4% (GTA1 w/ GPT-5). With the addition of Behavior Best-of-N, performance climbs even higher to 72.6%, surpassing human-level performance on OSWorld (~72%)!

Agent S3 also demonstrates strong zero-shot generalization! On WindowsAgentArena, accuracy rises from 50.2% using only Agent S3 to 56.6% by selecting from 3 rollouts. Similarly on AndroidWorld, performance improves from 68.1% to 71.6%

🛠️ Installation & Setup

Prerequisites

Single Monitor: Our agent is designed for single monitor screens
Security: The agent runs Python code to control your computer - use with care
Supported Platforms: Linux, Mac, and Windows

Installation

To install Agent S3 without cloning the repository, run

pip install gui-agents

If you would like to test Agent S3 while making changes, clone the repository and install using

pip install -e .

Don't forget to also brew install tesseract! Pytesseract requires this extra installation to work.

API Configuration

Option 1: Environment Variables

Add to your .bashrc (Linux) or .zshrc (MacOS):

export OPENAIAPIKEY=<YOURAPIKEY>
export ANTHROPICAPIKEY=<YOURANTHROPICAPI_KEY>
export HFTOKEN=<YOURHF_TOKEN>

Option 2: Python Script

import os
os.environ["OPENAIAPIKEY"] = "<YOURAPIKEY>"

Supported Models

We support Azure OpenAI, Anthropic, Gemini, Open Router, and vLLM inference. See models.md for details.

Grounding Models (Required)

For optimal performance, we recommend UI-TARS-1.5-7B hosted on Hugging Face Inference Endpoints or another provider. See Hugging Face Inference Endpoints for setup instructions.

🚀 Usage

⚡️ Recommended Setup:

For the best configuration, we recommend using OpenAI gpt-5-2025-08-07 as the main model, paired with UI-TARS-1.5-7B for grounding.

CLI

Note, this is running Agent S3, our improved agent, without bBoN.

Run Agent S3 with the required parameters:

agent_s \
    --provider openai \
    --model gpt-5-2025-08-07 \
    --ground_provider huggingface \
    --ground_url http://localhost:8080 \
    --ground_model ui-tars-1.5-7b \
    --grounding_width 1920 \
    --grounding_height 1080

Local Coding Environment (Optional)

For tasks that require code execution (e.g., data processing, file manipulation, system automation), you can enable the local coding environment:

agent_s \
    --provider openai \
    --model gpt-5-2025-08-07 \
    --ground_provider huggingface \
    --ground_url http://localhost:8080 \
    --ground_model ui-tars-1.5-7b \
    --grounding_width 1920 \
    --grounding_height 1080 \
    --enablelocalenv

⚠️ WARNING: The local coding environment executes arbitrary Python and Bash code locally on your machine. Only use this feature in trusted environments and with trusted inputs.

Required Parameters

--provider: Main generation model provider (e.g., openai, anthropic, etc.) - Default: "openai"
--model: Main generation model name (e.g., gpt-5-2025-08-07) - Default: "gpt-5-2025-08-07"
--ground_provider: The provider for the grounding model - Required
--ground_url: The URL of the grounding model - Required
--ground_model: The model name for the grounding model - Required
--grounding_width: Width of the output coordinate resolution from the grounding model - Required
--grounding_height: Height of the output coordinate resolution from the grounding model - Required

Optional Parameters

--model_temperature: The temperature to fix all model calls to (necessary to set to 1.0 for models like o3 but can be left blank for other models)

Grounding Model Dimensions

The grounding width and height should match the output coordinate resolution of your grounding model:

UI-TARS-1.5-7B: Use --groundingwidth 1920 --groundingheight 1080
UI-TARS-72B: Use --groundingwidth 1000 --groundingheight 1000

Optional Parameters

--model_url: Custom API URL for main generation model - Default: ""
--modelapikey: API key for main generation model - Default: ""
--groundapikey: API key for grounding model endpoint - Default: ""
--maxtrajectorylength: Maximum number of image turns to keep in trajectory - Default: 8
--enable_reflection: Enable reflection agent to assist the worker agent - Default: True
--enablelocalenv: Enable local coding environment for code execution (WARNING: Executes arbitrary code locally) - Default: False

Local Coding Environment Details

The local coding environment enables Agent S3 to execute Python and Bash code directly on your machine. This is particularly useful for:

Data Processing: Manipulating spreadsheets, CSV files, or databases
File Operations: Bulk file processing, content extraction, or file organization
System Automation: Configuration changes, system setup, or automation scripts
Code Development: Writing, editing, or executing code files
Text Processing: Document manipulation, content editing, or formatting

When enabled, the agent can use the callcodeagent action to execute code blocks for tasks that can be completed through programming rather than GUI interaction.

Requirements:

Python: The same Python interpreter used to run Agent S3 (automatically detected)
Bash: Available at /bin/bash (standard on macOS and Linux)
System Permissions: The agent runs with the same permissions as the user executing it

Security Considerations:

The local environment executes arbitrary code with the same permissions as the user running the agent
Only enable this feature in trusted environments
Be cautious when the agent generates code for system-level operations
Consider running in a sandboxed environment for untrusted tasks
Bash scripts are executed with a 30-second timeout to prevent hanging processes

`gui_agents` SDK

First, we import the necessary modules. AgentS3 is the main agent class for Agent S3. OSWorldACI is our grounding agent that translates agent actions into executable python code.

import pyautogui import io from guiagents.s3.agents.agents import AgentS3 from gui_agents.s3.agents.grounding import OSWorldACI from guiagents.s3.utils.localenv import LocalEnv  # Optional: for local coding environment
Load in your API keys.
from dotenv import load_dotenv
load_dotenv()
current_platform = "linux"  # "darwin", "windows"

Next, we define our engine parameters. engineparams is used for the main agent, and engineparamsforgrounding is for grounding. For engineparamsfor_grounding, we support custom endpoints like HuggingFace TGI, vLLM, and Open Router.

engine_params = {
  "engine_type": provider,
  "model": model,
  "baseurl": modelurl,           # Optional
  "apikey": modelapi_key,        # Optional
  "temperature": model_temperature # Optional
}
Load the grounding engine from a custom endpoint
groundprovider = "<yourground_provider>"
groundurl = "<yourground_url>"
groundmodel = "<yourground_model>"
groundapikey = "<yourgroundapi_key>"
Set grounding dimensions based on your model's output coordinate resolution
UI-TARS-1.5-7B: groundingwidth=1920, groundingheight=1080
UI-TARS-72B: groundingwidth=1000, groundingheight=1000
grounding_width = 1920  # Width of output coordinate resolution
grounding_height = 1080  # Height of output coordinate resolution
engineparamsfor_grounding = {   "enginetype": groundprovider,   "model": ground_model,   "baseurl": groundurl,   "apikey": groundapi_key,  # Optional   "groundingwidth": groundingwidth,   "groundingheight": groundingheight, }

Then, we define our grounding agent and Agent S3.

# Optional: Enable local coding environment
enablelocalenv = False  # Set to True to enable local code execution
localenv = LocalEnv() if enablelocal_env else None
grounding_agent = OSWorldACI(     env=localenv,  # Pass localenv for code execution capability     platform=current_platform,     engineparamsforgeneration=engineparams,     engineparamsforgrounding=engineparamsforgrounding,     width=1920,  # Optional: screen width     height=1080  # Optional: screen height )
agent = AgentS3(     engine_params,     grounding_agent,     platform=current_platform,     maxtrajectorylength=8,  # Optional: maximum image turns to keep     enable_reflection=True     # Optional: enable reflection agent )

Finally, let's query the agent!

# Get screenshot.
screenshot = pyautogui.screenshot()
buffered = io.BytesIO() 
screenshot.save(buffered, format="PNG")
screenshot_bytes = buffered.getvalue()
obs = {   "screenshot": screenshot_bytes, }
instruction = "Close VS Code" info, action = agent.predict(instruction=instruction, observation=obs)
exec(action[0])

Refer to guiagents/s3/cliapp.py for more details on how the inference loop works.

OSWorld

To deploy Agent S3 in OSWorld, follow the OSWorld Deployment instructions.

💬 Citations

If you find this codebase useful, please cite:

@misc{Agent-S3,
      title={The Unreasonable Effectiveness of Scaling Agents for Computer Use}, 
      author={Gonzalo Gonzalez-Pumariega and Vincent Tu and Chih-Lun Lee and Jiachen Yang and Ang Li and Xin Eric Wang},
      year={2025},
      eprint={2510.02250},
      archivePrefix={arXiv},
      primaryClass={cs.AI},
      url={https://arxiv.org/abs/2510.02250}, 
}
@misc{Agent-S2,       title={Agent S2: A Compositional Generalist-Specialist Framework for Computer Use Agents},        author={Saaket Agashe and Kyle Wong and Vincent Tu and Jiachen Yang and Ang Li and Xin Eric Wang},       year={2025},       eprint={2504.00906},       archivePrefix={arXiv},       primaryClass={cs.AI},       url={https://arxiv.org/abs/2504.00906},  }
@inproceedings{Agent-S,     title={{Agent S: An Open Agentic Framework that Uses Computers Like a Human}},     author={Saaket Agashe and Jiuzhou Han and Shuyu Gan and Jiachen Yang and Ang Li and Xin Eric Wang},     booktitle={International Conference on Learning Representations (ICLR)},     year={2025},     url={https://arxiv.org/abs/2410.08164} }