This is a 15-minute demo showcasing the use of Claude Desktop as an MCP Client leveraging the interop.io MCP Server and working directly with existing desktop applications on the io.Connect platfrom.
It follows our philosophy of “human-centered AI”, where the idea is not to replace experts or rebuild applications. Instead, it’s about providing AI with secure, governed access to your existing applications as “tools.” This allows the LLM to act as a powerful assistant that executes complex, multi-application workflows.
Critically, the professional user (e.g., wealth advisor) remains in full control. The user monitors every step, can change direction, correct mistakes, and has the final decision on all actions. This is essential because it places final authority and responsibility with the expert, not the LLM.
Stack:
-
AI Assistant UI (Claude Desktop): The user interacts with the LLM.
-
interop.io MCP Server: Securely exposes your existing applications and interop methods as “tools” that the LLM can invoke.
-
io.Connect: The platform that connects all your web or native applications.
-
Applications & Adapters: Your existing apps, including those integrated via io.Connect Adapters like PowerPoint and Outlook.
Key Takeaways
-
You can leverage LLMs without rebuilding your existing applications or processes.
-
The interop.io MCP Server allows you to securely expose your existing IO connect methods, adapters, and app functionalities as a catalog of “tools” for an AI to use.
-
This approach enables powerful, scalable automation of complex tasks while keeping the human user in full control of review and execution.
What do you think of using the MCP Server vs. direct tool calling? What other use cases or application “tools” would you want to expose to an AI assistant?