> ## Documentation Index
> Fetch the complete documentation index at: https://docs.cloud.cdata.com/llms.txt
> Use this file to discover all available pages before exploring further.

# CrewAI

> CrewAI is an open-source Python framework for building multi-agent systems. This page explains how to create a CrewAI agent to connect to the Connect AI MCP server.

## Prerequisites

Before you can configure and use CrewAI with Connect AI, you must do the following:

* Generate an [OAuth JWT bearer token](/en/API/Authentication-Embedded). Copy this down, as it acts as your password during authentication.
* Obtain an OpenAI API key: [https://platform.openai.com/](https://platform.openai.com/).
* Make sure you have Python >= 3.10 in order to use the CrewAI tools.

## Configure the Connect AI MCP Server

<Steps>
  <Step>
    Create a folder named `cdata-mcp-crew-agent`.
  </Step>

  <Step>
    Create a file with the extension `.env` in the `cdata-mcp-crew-agent` folder.
  </Step>

  <Step>
    Copy and paste the content below. Instead of MCP\_USERNAME and MCP\_PASSWORD, use the OAuth JWT token from the prerequisites to authenticate. Your OpenAI API key can be found at [https://platform.openai.com/](https://platform.openai.com/).

    ```bash theme={null}
    # CData MCP Server Configuration
    MCP_SERVER_URL="https://mcp.cloud.cdata.com/mcp"
    MCP_USERNAME="CONNECT_AI_EMAIL"
    MCP_PASSWORD="CONNECT_AI_PAT"
    OPENAI_API_KEY="OPEN_API_KEY"  # Your OPEN AI API Key
    ```
  </Step>
</Steps>

## Install the CrewAI Libraries

Run the following command in your terminal:

```bash theme={null}
pip install crewai requests python-dotenv
```

## Create and Run the CrewAI Agent

<Steps>
  <Step>
    Create a file called `crew-agent.py`. This is the CrewAI agent.
  </Step>

  <Step>
    Copy and paste the following, defining the agent's tasks as desired:

    <Note>
      **Note:** For Connect AI Embed, you do not need the `MCP_USERNAME` or `MCP_PASSWORD`. Enter your OAuth JWT credentials in `auth_header`.
    </Note>

    ```python expandable theme={null}
    import os
    import base64
    import requests
    from dotenv import load_dotenv
    from crewai import Agent, Task, Crew
    # Load environment variables
    load_dotenv()
    # MCP Server configuration
    MCP_SERVER_URL = os.getenv('MCP_SERVER_URL', 'https://mcp.cloud.cdata.com/mcp')
    MCP_USERNAME = os.getenv('MCP_USERNAME', '')
    MCP_PASSWORD = os.getenv('MCP_PASSWORD', '')
    # Create Basic Auth header
    auth_header = {}
    if MCP_USERNAME and MCP_PASSWORD:
        credentials = f"{MCP_USERNAME}:{MCP_PASSWORD}"
        auth_header = {
            "Authorization": f"Basic {base64.b64encode(credentials.encode()).decode()}"
        }
    # MCP tool invocation function
    def call_mcp_tool(tool_name, input_data):
        payload = {
            "tool": tool_name,
            "input": input_data
        }
        try:
            response = requests.post(
                MCP_SERVER_URL,
                json=payload,
                headers=auth_header,
                timeout=10
            )
            response.raise_for_status()
            return response.json()
        except Exception as e:
            return {"error": str(e)}
    # Define CrewAI agent
    class MCPQueryAgent(Agent):
        def __init__(self):
            super().__init__(
                name="cdata_query_assistant",
                role="Data Query Assistant",
                goal="Help users explore and query Connect AI databases",
                backstory="You are a helpful assistant trained to interact with Connect AI via MCP tools. You understand databases, schemas, and SQL queries, and you guide users through their data exploration journey."
            )
        def get_catalogs(self):
            return call_mcp_tool("getCatalogs", {})
        def get_schemas(self, catalog):
            return call_mcp_tool("getSchemas", {"catalog": catalog})
        def get_tables(self, catalog, schema):
            return call_mcp_tool("getTables", {"catalog": catalog, "schema": schema})
        def get_columns(self, catalog, schema, table):
            return call_mcp_tool("getColumns", {
                "catalog": catalog,
                "schema": schema,
                "table": table
            })
        def query_data(self, catalog, query):
            return call_mcp_tool("queryData", {
                "catalog": catalog,
                "query": query
            })
    # Instantiate agent
    agent = MCPQueryAgent()
    # Define tasks
    task1 = Task(
        agent=agent,
        description="List the top 10 available catalogs in Connect AI",
        expected_output="Catalog list",
        output_function=lambda agent: agent.get_catalogs()
    )
    # Create and run crew
    crew = Crew(
        agents=[agent],
        tasks=[task1]
    )
    if __name__ == "__main__":
        results = crew.kickoff()
        print("\n=== Final Output ===")
        for result in results:
            print(result)
    ```
  </Step>

  <Step>
    Run the following command in the terminal. The output displays the results of the task:

    ```bash theme={null}
    python crew-agent.py
    ```

    <Frame>
      <img src="https://mintcdn.com/cdata/dJhcR4yZCQgD8aFQ/en/images/crewai_client_output.png?fit=max&auto=format&n=dJhcR4yZCQgD8aFQ&q=85&s=838de0b425c9abcc6ffc08bb05d05cc8" alt="CrewAI Client Output" width="1920" height="1363" data-path="en/images/crewai_client_output.png" />
    </Frame>
  </Step>
</Steps>