monday.com has introduced MCP Apps, an extension to the Model Context Protocol, allowing AI tools to incorporate interactive UI elements within chat interfaces. This capability enables users to interact with their project data directly in conversations, without switching contexts or needing to copy and paste between windows. For instance, when a user inquires about their project status, they can see a battery widget that visually represents task statuses. By clicking on segments of this widget, they can access detailed tables and context with minimal effort. The development of MCP Apps began with a simple prototype, embedding an existing monday.com dashboard into the Goose chat application. This approach allowed the team to gather valuable insights quickly, identifying challenges with authentication and state management. The team established principles guiding their design, such as focusing on where UI can enhance communication rather than overwhelm users with unnecessary details. They concluded that interactions should be ephemeral—supporting immediate questions without expecting users to navigate a full application. Key design rules emerged from this process. For example, instead of requiring users to describe what they see, the system allows for direct interactions, like clicking on a status segment to filter items. The agent handles data flow, meaning users can make requests in natural language without needing to configure dropdowns or settings. This makes the chat experience more fluid and responsive. Importantly, embedding full applications was deemed impractical, as it disrupts the chat’s efficiency. Instead, the focus remains on leveraging existing contextual information from connected tools to streamline interactions.