I'd like to suggest the development of a dedicated desktop application for Kagi Intelligence that supports the Model Context Protocol (MCP). I'm aware that Kagi already has an MCP server (https://github.com/kagisearch/kagimcp), but my suggestion is specifically for a Kagi Intelligence desktop app that would act as an MCP client to connect KI to other MCP servers.
This would allow Kagi's AI assistant to seamlessly integrate with our desktop workflows and interact with other applications on our devices.
The desktop client would provide a native experience across operating systems (Windows, macOS, Linux) while maintaining all the functionality of the web version. By acting as an MCP client, it would allow Kagi Intelligence to access relevant context from other applications I'm using - such as being aware of content I'm viewing, documents I'm editing, or code I'm writing.
This would dramatically improve productivity for users by eliminating the need to switch contexts or manually copy information between applications and the browser. It would also preserve privacy by keeping sensitive context local rather than requiring uploads to web interfaces.
I envision several powerful use cases, assuming the MCP gets broadly adopted in other applications:
Contextual research assistance: While working on a document in Word or Google Docs, I could invoke Kagi Intelligence through a keyboard shortcut, and it would automatically understand what I'm working on. I could ask "Find me more sources about this topic" or "Suggest improvements to this paragraph" without having to copy-paste or explain what I'm working on.
Programming support: When coding in VS Code or other IDEs, the desktop client could understand my codebase through MCP. I could ask "What's wrong with this function?" or "How do I implement feature X in this codebase?" and get contextually relevant assistance.
System-wide search and retrieval: The client could leverage MCP to search across my local files and applications while using Kagi's superior search capabilities, creating a unified interface for both local and web information.
Contextual workflows across applications: I might research in my browser, take notes in Notion, and then draft an email - the desktop client with MCP could follow this context chain, maintaining awareness of the entire workflow.
Similar implementations exist in tools like GitHub Copilot (recently got support for acting as a MCP client) and Claude, but none combine Kagi's superior search capabilities with MCP's contextual awareness across the entire desktop environment.
This would extend Kagi Intelligence from being a web-based AI assistant to becoming a truly integrated productivity partner that moves seamlessly between applications while maintaining context and leveraging Kagi's exceptional information retrieval capabilities.
