mcp-aoai-web-browsing

MCP Server & Client implementation for using Azure OpenAI

This repository provides a minimal server/client application that leverages the Model Context Protocol (MCP) with Azure OpenAI. The MCP server is built using FastMCP, and Playwright is used for web application testing. The MCP response regarding available tools is converted into the OpenAI function calling format via a customized MCP-LLM Bridge implementation. The server object is directly passed into the bridge to ensure a stable connection.

The project utilizes uv for Python dependency management. Configuration involves setting up Azure OpenAI credentials in the .env file. The chatgui.py script demonstrates a client launching a browser to navigate to a specified URL. The repository also provides guidance on using stdio and JSON-RPC, along with examples of tool descriptions and uv commands. It includes tips for debugging with Visual Studio Code and terminating Python processes.