Download
MolmoWeb is an open-source multimodal web agent designed to autonomously navigate and interact with web browsers using vision-language models, representing a significant step toward fully agentic AI systems that can operate in real-world digital environments. The system takes natural language instructions and translates them into sequences of browser actions such as clicking, typing, scrolling, and navigating, effectively performing tasks on behalf of the user. Unlike traditional automation tools that rely on structured HTML parsing or predefined APIs, MolmoWeb operates directly from screenshots of web pages, interpreting visual content in the same way a human user would. This approach allows it to generalize across different websites without requiring site-specific integrations, making it highly adaptable to diverse web environments.
Features
- Autonomous browser control through natural language instructions
- Vision-based interaction using screenshots instead of HTML parsing
- Execution of actions such as clicking, typing, scrolling, and navigation
- Open-source models, datasets, and evaluation pipeline for reproducibility
- Multi-step reasoning loop combining perception, decision, and action
- Self-hosted deployment with full control over infrastructure and data
Project Samples
