Anthropic Expands MCP Using New UI Framework
Anthropic has significantly enhanced its MCP protocol with the introduction of a new user interface (UI) framework. This advancement aims to create a more interactive app-like experience within chat interfaces. With this release, third-party applications can seamlessly integrate into the MCP ecosystem, allowing users to interact in innovative ways directly from their chat sessions.
Introduction of MCP Apps
The MCP protocol, established by Anthropic, serves as a standard for communication between AI models and various applications. In December, Anthropic donated this protocol to the Agentic AI Foundation, paving the way for broader adoption.
The newly released extension enables MCP servers to deliver an interactive experience, enhancing the functionality of chatbots like Claude. Other chatbot providers can adopt this open protocol, fostering further development of interactive applications.
Current Support for MCP Apps
Several applications currently support the MCP framework:
- Goose
- Visual Studio Code (for Insiders)
- OpenAI’s ChatGPT (coming soon)
Partnerships and Features
Anthropic has partnered with various organizations to implement the MCP Apps. Notable partners include:
- Amplitude
- Asana
- Box
- Canva
- Clay
- Figma
- Hex
- monday.com
- Slack
The Box MCP app, for example, allows users to search and preview documents within chat. Users can also ask questions about these documents, making the interaction more dynamic. Similarly, the Slack MCP app enables users to draft messages, edit them, and post directly to Slack.
Practical Applications
MCP Apps allow users to engage in various scenarios, such as:
- Interactive data exploration using dashboards
- Configuration wizards
- Document reviews
- Real-time monitoring
Technical Insights
The foundation of MCP Apps employs UI metadata and resources, specifically HTML and JavaScript, to create interactive experiences. This transition to a more visual and engaging interaction method aligns with the current need for enhanced usability in AI chat systems.
Anthropic acknowledges that it is not alone in this endeavor. OpenAI’s Apps SDK also provides a similar experience, illustrating the growing trend of integrating apps within chat interfaces.
Security Measures
On the security front, Anthropic has implemented stringent measures to safeguard user data. These include:
- Sandboxed iframes with restricted permissions
- Review capability for HTML content before rendering
- Auditable messages between UI and host
- Explicit user approval required for UI-initiated tool calls
Future Prospects
The advancements in MCP Apps present exciting opportunities for developers. As more applications and functionalities are created, users can look forward to a richer experience within chat-centric AI tools.
Anthropic’s initiative promises to revolutionize the interaction model of AI applications, making it accessible and effective for various enterprises.
