MCP Apps is the future of MCP
What is MCP Apps?
With MCP, Anthropic standardized how AI models connect to external tools. Suddenly, models could access real-time data and execute actions — and that completely changed our perception of what AI could do. But MCP had a limitation: tools could only respond with text. Chatbots got much smarter, yet their answers were still stuck in plain text. MCP is fully supported by LocalIntelligence 1.5.
That’s why, even amazed by what chatbots could do, users still turned to their web browsers for shopping, learning, and countless other tasks, because the visual experience just wasn’t there.
MCP Apps solves this by letting tools respond with rich HTML/CSS web content. This closes the gap between chat and web UX. Chatbots can now deliver experiences that rival native web apps — animations, dynamic charts, interactive components — directly in the conversation.
The momentum is real. OpenAI and Anthropic already use the emerging MCP Apps standard (SEP-1865) to render dynamic content in their products, and Shopify is among the early adopters pushing for its adoption.
Announcing LocalIntelligence 2.0
Today I’m happy to announce that I’ve extended LocalIntelligence, my Ollama front-end app, to support MCP Apps. This means that it one of the first native macOS apps to do so. That makes it a great environment for anyone who wants to build and test MCP App servers locally.
LocalIntelligence still needs a little polish before I share it publicly, but it should be available for download soon. In the meantime, here’s a screenshot of LocalIntelligence running a simple MCP tool right inside the chat.