agent-vision-toolkit

Agent Vision Toolkit is an open-source collection of visual analysis tools and agent skills that gives text-only AI models access to image understanding capabilities. It supports image question answering, screenshot interpretation, optical character recognition, and visual element identification. Specialized workflows help agents reconstruct web interfaces from screenshots and compare rendered results against references. The toolkit provides image cropping, object grounding, tracing, and visual difference analysis. Its companion skill guides agents through selecting appropriate tools and verifying their results. A local proxy and optional plugins allow pasted images to work transparently with supported coding assistants. Integrations include Codex, Claude Code, Pi, Oh My Pi, and OpenCode. The software enables multimodal workflows without requiring a natively multimodal primary model.