An open source alternative to MultiOn's visual grounding engine. It maps screenshots to interactive page elements to enable multimodal LLMs to navigate the web without complex DOM parsing.
Tech Stack
Jupyter NotebookAI, ML & Data
Why tarsier?
• Simplifies complex DOM trees
• Works with any vision LLM
• Reduces token consumption
Limitations
• High latency for screenshots
• Experimental status
• Requires vision-capable models
3/5/2026
Last Update
121
Forks
17
Issues
MIT
License
Financial Leak Detected
Stop the "SaaS Tax"
Your team could be burning cash. Switching to tarsier instantly boosts your runway.
