Integrating LlamaIndex with Azure Data Lake Storage Data via CData Connect AI

LlamaIndex is a data framework for building LLM applications — agents, RAG pipelines, and structured workflows that reason over external data. By integrating LlamaIndex with CData Connect AI through the built-in MCP Server, your agents can discover and query live Azure Data Lake Storage data as native tools without writing custom connectors.

CData Connect AI offers a secure, low-code environment to connect Azure Data Lake Storage and other data sources, removing the need for complex ETL and enabling seamless automation across business applications with live data.

This article outlines how to configure Azure Data Lake Storage connectivity in CData Connect AI, register the MCP server with LlamaIndex, and build a ReAct agent that queries Azure Data Lake Storage data in real time.

Prerequisites

• An account in CData Connect AI
• Python version 3.10 or higher, to install the LlamaIndex packages
• Generate and save an OpenAI API key
• Install Visual Studio Code in your system

Step 1: Configure Azure Data Lake Storage Connectivity for LlamaIndex

Before LlamaIndex can access Azure Data Lake Storage, a Azure Data Lake Storage connection must be created in CData Connect AI. This connection is then exposed to LlamaIndex through the remote MCP server.

1. Log in to Connect AI, click Sources, and then click + Add Connection
2. From the available data sources, choose Azure Data Lake Storage
3. Enter the necessary authentication properties to connect to Azure Data Lake Storage

   Authenticating to a Gen 1 DataLakeStore Account

   Gen 1 uses OAuth 2.0 in Entra ID (formerly Azure AD) for authentication.

   For this, an Active Directory web application is required. You can create one as follows:

   1. Sign in to your Azure Account through the .
   2. Select "Entra ID" (formerly Azure AD).
   3. Select "App registrations".
   4. Select "New application registration".
   5. Provide a name and URL for the application. Select Web app for the type of application you want to create.
   6. Select "Required permissions" and change the required permissions for this app. At a minimum, "Azure Data Lake" and "Windows Azure Service Management API" are required.
   7. Select "Key" and generate a new key. Add a description, a duration, and take note of the generated key. You won't be able to see it again.

   To authenticate against a Gen 1 DataLakeStore account, the following properties are required:

   • Schema: Set this to ADLSGen1.
   • Account: Set this to the name of the account.
   • OAuthClientId: Set this to the application Id of the app you created.
   • OAuthClientSecret: Set this to the key generated for the app you created.
   • TenantId: Set this to the tenant Id. See the property for more information on how to acquire this.
   • Directory: Set this to the path which will be used to store the replicated file. If not specified, the root directory will be used.

   Authenticating to a Gen 2 DataLakeStore Account

   To authenticate against a Gen 2 DataLakeStore account, the following properties are required:

   • Schema: Set this to ADLSGen2.
   • Account: Set this to the name of the account.
   • FileSystem: Set this to the file system which will be used for this account.
   • AccessKey: Set this to the access key which will be used to authenticate the calls to the API. See the property for more information on how to acquire this.
   • Directory: Set this to the path which will be used to store the replicated file. If not specified, the root directory will be used.
   
4. Click Save & Test
5. Once authenticated, open the Permissions tab in the Azure Data Lake Storage connection and configure user-based permissions as required

Generate a Personal Access Token (PAT)

LlamaIndex authenticates to Connect AI using an account email and a Personal Access Token (PAT). Creating separate PATs for each integration is recommended to maintain access control granularity.

1. In Connect AI, select the Gear icon in the top-right to open Settings
2. Under Access Tokens, select Create PAT
3. Provide a descriptive name for the token and select Create
4. Copy the token and store it securely. The PAT will only be visible during creation

With the Azure Data Lake Storage connection configured and a PAT generated, LlamaIndex is prepared to connect to Azure Data Lake Storage data through the CData MCP server.

Step 2: Connect to the MCP server in LlamaIndex

To connect LlamaIndex with CData Connect AI Remote MCP Server and use OpenAI for reasoning, configure your MCP server endpoint and authentication in a

config.py

file. These values let LlamaIndex’s MCP tool spec call the MCP server tools, while OpenAI handles the natural language reasoning.

1. Create a folder for the LlamaIndex MCP project
2. Create two Python files within the folder:

   config.py

   and

   llamaindex_agent.py

3. In

   config.py

   , define your MCP server URL and your Base64-encoded CData Connect AI email and PAT (obtained in the prerequisites):

   class Config:
   
         MCP_BASE_URL = "https://mcp.cloud.cdata.com/mcp"   # MCP Server URL
         MCP_AUTH     = "base64encoded(EMAIL:PAT)"          # Base64 encoded Connect AI Email:PAT
   

   Note: You can create the base64 encoded version of MCP_AUTH using any Base64 encoding tool.

4. In

   llamaindex_agent.py

   , wire up the MCP tool spec and a ReAct agent:

   """
   Integrates a LlamaIndex ReAct agent with the CData Connect AI MCP server.
   The script discovers MCP tools, wraps them as LlamaIndex tools, and runs an
   agent loop driven by OpenAI for reasoning.
   """
   
   import asyncio
   from llama_index.tools.mcp import BasicMCPClient, McpToolSpec
   from llama_index.core.agent.workflow import ReActAgent
   from llama_index.llms.openai import OpenAI
   from config import Config
   
   async def main():
   
       # Initialize the MCP client pointed at Connect AI
       mcp_client = BasicMCPClient(
           Config.MCP_BASE_URL,
           headers={"Authorization": f"Basic {Config.MCP_AUTH}"},
       )
   
       # Discover tools the MCP server exposes (getCatalogs, queryData, etc.)
       tool_spec = McpToolSpec(client=mcp_client)
       tools = await tool_spec.to_tool_list_async()
       print("Discovered MCP tools:", [t.metadata.name for t in tools])
   
       # Configure the LLM that drives the ReAct loop
       llm = OpenAI(
           model="gpt-4o",
           temperature=0.2,
           api_key="YOUR_OPENAI_API_KEY",  # https://platform.openai.com/
       )
   
       # Build the agent with the MCP-backed tools
       agent = ReActAgent(tools=tools, llm=llm)
   
       user_prompt = "How many tables are available in ADLS1?"  # Change as needed
       print(f"
   User prompt: {user_prompt}")
   
       response = await agent.run(user_prompt)
   
       print("Agent final response:", response)
   
   if __name__ == "__main__":
       asyncio.run(main())
   

Step 3: Install the LlamaIndex packages

Since this workflow uses LlamaIndex together with the CData Connect AI MCP server and OpenAI for reasoning, install the required Python packages.

Run the following command in your project terminal:

pip install llama-index llama-index-tools-mcp llama-index-llms-openai

Step 4: Prompt Azure Data Lake Storage using LlamaIndex (via the MCP server)

1. When the installation finishes, run

   python llamaindex_agent.py

   to execute the script
2. The script connects to the MCP server and discovers the CData Connect AI MCP tools available for querying your connected data
3. Supply a prompt (e.g., "How many tables are available in Azure Data Lake Storage?")
4. The agent reasons over the available tools, calls

   queryData

   against Azure Data Lake Storage, and responds with the result

Get CData Connect AI

To get live data access to hundreds of SaaS, Big Data, and NoSQL sources directly from your cloud applications, try CData Connect AI today!