Introduction

Setting up MCPs whether on your local machine or using remote ones is always a challenge in itself - we simplify this. Editing what tools your agent can see, or modifying how your agent perceives them forces you to edit the access the source code. We overcome this challenge by providing an easy and intuitive UI to manage all your MCP deployments and usage in a single place.

What it does

MCProxy is an intelligent middle layer for MCP servers that: • Dynamically personalizes tool descriptions for specific AI agents (e.g., translates language or rewrites descriptions for different LLMs like Llama or Gemini). • Provides a central hub to manage all connected MCP servers. • Allows easy enabling/disabling of specific tools for agents. • Enables full auditing of traffic between agents and MCP servers. (Not implemented with demo, but our solution allows to add this feature )

How we built it

We built a Web App that is a middleware solution that sits between AI agents and MCP servers, intercepting and adapting communications on the fly.

Accomplishments that we're proud of

We've successfully developed a working prototype that demonstrates MCProxy's core capabilities in real-time personalization and centralized control for two MCP servers

Built With

Share this project:

Updates