Airflow + Microsoft Copilot Studio: Complete MCP Integration

Airflow is a MCP server that A MCP Server that connects to Apache Airflow using official python client..

When integrated with Microsoft Copilot Studio, you can:

• List all available DAGs in Airflow
• Manually trigger a DAG run
• Get DAG run history and status

This guide provides step-by-step instructions to set up Airflow in Microsoft Copilot Studio, including configuration, examples, and troubleshooting.

What You'll Achieve

After completing this setup:

• Airflow will be fully integrated and operational
• You can use Airflow tools directly in Microsoft Copilot Studio
• All Airflow capabilities will be available for your workflows
• Access to 5 different tools

Prerequisites

Before starting, ensure you have:

• Airflow webserver host URL
• Airflow authentication username
• Airflow authentication password
• Microsoft Copilot Studio installed and configured
• Compatible operating system (Web Browser, Microsoft 365)

Installation

Step 1: Install Airflow

Configuration

Step 2: Configure Microsoft Copilot Studio

1. Enable Generative Orchestration

Generative Orchestration must be enabled in your Copilot Studio environment to use MCP

Visit: https://learn.microsoft.com/en-us/microsoft-copilot-studio/agent-extend-action-mcp

Note: This is a prerequisite - MCP will not work without it

2. Access Copilot Studio

Log in to Microsoft Copilot Studio with appropriate permissions to create and configure agents

Visit: https://copilotstudio.microsoft.com

Note: Requires organizational access or trial subscription

3. Navigate to Connectors

In Copilot Studio, go to the Connectors section where you can add new integrations

Note: Location may vary based on interface updates

4. Create MCP Connector

Click "Add Connector" and select "Model Context Protocol (MCP)" from the available connector types

Note: Use the YAML schema template provided by Copilot Studio

5. Configure {server_name} Connection

Fill in the YAML schema with {server_name} server details including command, arguments, and environment variables

airflow

Note: Copilot Studio dynamically reflects changes as tools and resources are updated on the MCP server

6. Add Tools to Agent

Once connected, add the MCP server tools and resources to your specific agent through the Copilot Studio UI

Note: Currently supports Tools and Resources (Prompts not yet supported)

7. Test Agent Capabilities

Test your agent to verify it can access and use the {server_name} tools correctly

Configuration Details

Connect MCP server to Microsoft Copilot Studio:

Step 1: Create YAML Connector Schema

In Copilot Studio, navigate to the connectors section and create a new MCP connector using the YAML schema template:

# MCP Connector Configuration for Airflow
name: Airflow-mcp-connector
description: A MCP Server that connects to Apache Airflow using official python client.

mcp_server:
  command: airflow
  args: 
  env: "AIRFLOW_HOST": "http://localhost:8080",
        "AIRFLOW_USERNAME": "admin",
        "AIRFLOW_PASSWORD": "admin_password",
        "AIRFLOW_READ_ONLY": "true"

Step 2: Add to Agent

Once connected, add the MCP server's tools and resources to your Copilot Studio agent through the UI.

Examples

Once configured, you can use Airflow in Microsoft Copilot Studio:

Pipeline Monitoring

undefined

Ask Microsoft Copilot Studio: "Check status of all DAGs and recent failures"

Expected Result: undefined

Manual DAG Execution

undefined

Ask Microsoft Copilot Studio: "Trigger monthly_report DAG with date parameter"

Expected Result: undefined

Troubleshooting Failed Tasks

undefined

Ask Microsoft Copilot Studio: "Show failed tasks in data_pipeline and their error logs"

Expected Result: undefined

Testing Your Setup

1. Launch Microsoft Copilot Studio
2. Verify Airflow is available in the tools list
3. Test basic Airflow functionality

Troubleshooting

Common Issues

Generative Orchestration Not Available

Symptoms: MCP option not visible, Cannot add MCP connectors

Cause: Generative Orchestration is not enabled for your environment

Solution:

• Contact your Microsoft 365 administrator
• Enable Generative Orchestration in tenant settings
• Verify your subscription includes Copilot Studio with MCP support
• Check Microsoft Learn documentation for enablement steps

MCP Connector Connection Failed

Symptoms: Connector shows "Disconnected" status, Tools not available to agent

Cause: MCP server not accessible or configuration error

Solution:

• Verify the MCP server command is correct and accessible
• Check all required environment variables are set
• Test the MCP server independently outside Copilot Studio
• Review YAML schema for syntax errors
• Check network connectivity and firewall rules

Tools Not Appearing in Agent

Symptoms: Connector connected but tools missing, Agent cannot invoke MCP tools

Cause: Tools not added to agent or synchronization issue

Solution:

• Manually add tools to agent through Copilot Studio UI
• Refresh connector to sync latest tools from MCP server
• Verify tools are published on the MCP server
• Check agent configuration includes the MCP connector

MCP Server Tools Update Not Reflected

Symptoms: New tools not appearing, Old tools still visible after removal

Cause: Copilot Studio cache or sync delay

Solution:

• Disconnect and reconnect the MCP connector
• Wait a few minutes for dynamic sync to occur
• Manually refresh the connector in Copilot Studio
• Restart the MCP server if possible

Airflow not appearing in Microsoft Copilot Studio

Symptoms: Server not listed, Tools not available

Cause: Configuration or installation issue

Solution:

• Verify configuration syntax
• Check Airflow installation
• Restart Microsoft Copilot Studio
• Check logs for error messages

Next Steps

Now that Airflow is integrated with Microsoft Copilot Studio:

• Explore all Airflow capabilities through Microsoft Copilot Studio
• Check out other MCP servers that work with Microsoft Copilot Studio
• Join the MCP community for tips and support
• Consider contributing to Airflow development

Need Help?

• Search for Airflow documentation
• Check the Microsoft Copilot Studio MCP guide
• Join the MCP community discussions