A new open-source iOS app called VisionClaude connects Anthropic's Claude vision API to your iPhone camera or Meta Ray-Ban smart glasses, creating a hands-free AI assistant you control entirely with your voice.
The app captures frames from your iPhone's camera at 1080p (or 720p from Ray-Ban glasses) and sends them to Claude for real-time analysis. You speak a question, Claude sees what you're looking at, and you hear a response through ElevenLabs text-to-speech. A Node.js gateway server running on your Mac coordinates everything between the iOS app, Claude's API, and any MCP tool servers you've configured.
The practical upside is the MCP integration. Because VisionClaude connects to Model Context Protocol servers (the same standard Claude Desktop uses), it can trigger actions like sending emails, checking your calendar, or posting messages while you're walking around with your glasses on. There's also a "skills" system that loads custom instruction files to shape how Claude responds in different contexts.
Setup isn't trivial. You need API keys for both Anthropic and ElevenLabs, a Mac running the gateway server, and for the Ray-Ban integration, Meta's DAT SDK and a Bluetooth connection. This is a developer-oriented project, not a polished consumer app.
Still, it's one of the more interesting wearable AI experiments built on Claude. The combination of real-time vision, voice I/O, and tool access through MCP makes it a useful reference for anyone building hands-free AI interfaces on Apple hardware.