Agent S2 (Simular)
Agent S2 (Simular) is an open-source framework for building GUI agents that autonomously interact with computers through an Agent-Computer Interface. It solves the problem of automating complex computer tasks by learning from past experiences and using a compositional generalist-specialist approach. It is for developers and researchers interested in AI-driven computer use agents.
🎬 Agent S3 - Near human-level computer use agent · Simular
✨ Key features
- State-of-the-art computer use agent framework
- Supports Linux, Mac, and Windows platforms
- Compatible with multiple LLM providers (OpenAI, Anthropic, Gemini, etc.)
- Optional local coding environment for code execution
- Reflection agent to assist the worker agent
- Grounding models for accurate screen interaction
🎯 Use cases
- Automating GUI tasks like closing applications or file operations
- Data processing and spreadsheet manipulation
- System automation and configuration changes
- Code development and execution
- Text processing and document editing
📦 Installation
🧰 Requirements: Requires Python, a single monitor, API keys for OpenAI/Anthropic/Hugging Face, and a grounding model (e.g., UI-TARS-1.5-7B).
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.
🚀 Usage
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
⚠️ Good to know
The agent is designed for single monitor screens and executes Python code to control your computer, so use with care; the local coding environment executes arbitrary code and should only be used in trusted environments.
❓ FAQ
What are the supported platforms?
Linux, Mac, and Windows are supported.
Which models are supported?
We support Azure OpenAI, Anthropic, Gemini, Open Router, and vLLM inference.
What grounding model is recommended?
For optimal performance, we recommend UI-TARS-1.5-7B hosted on Hugging Face Inference Endpoints.
How do I set up API keys?
You can set environment variables OPENAI_API_KEY, ANTHROPIC_API_KEY, and HF_TOKEN, or set them in a Python script using os.environ.
📊 Repository
🤖 Overview, features, install steps and FAQ were generated from the project's README on Sep 4, 2026. Always check the original source before running commands.