MCP Foundry DevDocs

This repository contains tools for document processing and AI agent integration:

1. Azure AI Agent MCP Server: A Model Context Protocol (MCP) server that integrates with Azure AI Agent Service
2. OpenAI API Specification Processor: A utility to prepare OpenAI's API documentation for Retrieval Augmented Generation (RAG)

Azure AI Agent MCP Server

An MCP server that connects to Azure AI Agent Service, allowing VS Code extensions and other MCP clients to interact with Azure AI Agents.

Features

• Connect to any agent in your Azure AI project
• Query a default agent with a simpler interface
• List all available agents
• Format responses with proper Markdown including citations

Setup

1. Create a virtual environment:

   python -m venv .venv
   source .venv/bin/activate  # On Windows: .venv\Scripts\activate
   

2. Install dependencies:

   pip install -r requirements.txt
   

3. Configure environment variables in .env:

   # Required for Azure AI Agent MCP Server
   PROJECT_CONNECTION_STRING=your_project_connection_string
   DEFAULT_AGENT_ID=your_default_agent_id
   

4. The connection string format should be:

   [region].api.azureml.ms;[tenant_id];[project_name];[workspace_name]
   

Running the MCP Server

Start the server:

python -m azure_agent_mcp_server

Configuring in VS Code

Add the server to your VS Code settings.json:

"mcp": {
  "servers": {
    "azure-agent": {
      "command": "/path/to/your/.venv/bin/python",
      "args": [
        "-m",
        "azure_agent_mcp_server"
      ],
      "cwd": "/path/to/your/src",
      "env": {
        "PYTHONPATH": "/path/to/your/src",
        "PROJECT_CONNECTION_STRING": "your_connection_string",
        "DEFAULT_AGENT_ID": "your_default_agent_id"
      }
    }
  }
}

Available Tools

The MCP server provides these tools:

• connect_agent: Connect to a specific Azure AI Agent using its ID
• query_default_agent: Query the default agent specified in your configuration
• list_agents: List all available agents in your Azure AI project

OpenAI API Specification Processor

A Python utility to fetch, process, and index the OpenAI API specification for RAG applications.

Overview

This tool:

• Fetches the OpenAI API specification from GitHub
• Processes it into searchable chunks
• Uploads the chunks to Azure Blob Storage
• Creates and configures an Azure AI Search index

Setup

Configure environment variables in .env (in addition to the Azure AI Agent variables):

# Azure Storage Account settings
AZURE_STORAGE_CONNECTION_STRING=your_storage_connection_string
AZURE_STORAGE_ACCOUNT_NAME=your_storage_account_name
AZURE_STORAGE_CONTAINER_NAME=your_container_name

# Azure AI Search settings
AZURE_SEARCH_SERVICE_ENDPOINT=https://your-service-name.search.windows.net
AZURE_SEARCH_API_KEY=your_search_api_key
AZURE_SEARCH_INDEX_NAME=your_index_name

# Optional OpenAI API key for embeddings
OPENAI_API_KEY=your_openai_api_key

Usage

Run the processor:

python src/openai_spec_processor.py

Command Line Options

• --fetch-only: Only fetch the API spec
• --process-only: Only process the API spec (no upload)
• --upload-only: Only upload processed documents
• --index-only: Only create/update the search index
• --chunk-size: Size of content chunks (default: 1000)
• --chunk-overlap: Overlap between chunks (default: 200)
• --skip-embeddings: Skip generating embeddings
• --test-query: Run a test query against the index

Example:

python src/openai_spec_processor.py --test-query "How do I use the completions API?"