Skip to main content
Skip table of contents

KADA MCP for Copilot Studio (Beta)

Kada provides a MCP integration for Copilot Studio.

You must request access to the Kada MCP beta. Reach out to your Kada account manager for access.

Prerequisites

  1. You must have access to Copilot Studio and licensed for Microsoft 365 Copilot

  2. Kada MCP API key

  3. Kada MCP Server URL endpoint (e.g., https://demo.kada.ai/mcp)

  4. Kada Agent knowledge files

    1. kada_default_columns.txt

    2. lineage_rules.txt

    3. solr_kada_search_roles.txt

Step 1: Adding the MCP Server as a Copilot Agent Tool

  1. Log into Copilot studio - copilotstudio.microsoft.com/

  2. Click on Tools in the navigation pane

  3. Click Add a tool

  4. Select + New tool

  5. Select Model Context Protocol

Step 2: Add in MCP Server Information

Provide the following values in the MCP configuration form

Field

Example

Description

Server name

KADA MCP SERVER

Short, clear name for the MCP server

Server description

KADA MCP endpoint for Copilot integration

Detailed description - the agent uses this to decide when to call the MCP server

Server URL

https://<your-deployed-server-url>/mcp

Your MCP server endpoint provided by Kada (must include /mcp path)

Step 3: Configure MCP Authentication

  1. Under Authentication, select API key

  2. For Type, select Header

  3. For Header name, enter: x-api-key

  4. Click Create

Step 4: Create MCP Connection

After creating the server configuration, you'll see the "Add tool" page:

  1. The server appears but shows "Not connected"

  2. Click Create new connection




  3. In the connection dialog:

    • Authentication Type is pre-selected as "API Key"

    • Enter your API key provided by Kada

  4. Click Create


Step 5: Add to MCP to your Agent

  1. After creating the connection, you'll return to the "Add tool" page

  2. Select your newly created connection from the Connection dropdown

  3. Click Add and configure


  4. The Kada MCP server is now available to your agent.

Step 6: Configuring your agent

  1. Click on Agents in the navigation pane

  2. Click Add agent

  3. In the agent creation interface, click on the Configure tab

  1. Configure Agent Details:

    1. Name: Enter a descriptive name for your agent (e.g., "K.AI")

    2. Description: Add a brief description of what this agent does.

    3. Instructions: Paste the instructions provided by KADA in the Instructions field.

    4. Logo: <TBC>

  2. Click Create

Step 7: Configuring your Agent

  1. Navigate to your agent from Agents in the navigation pane.

  2. In the Overview tab go to the Instructions section. Click Edit

  3. Enter the following instructions

    Always search Kada by calling the KADA MCP SERVER.
    When you construct the solr query always use 'kada_default_columns.txt' and 'solr_kada_search_roles.txt' to use the correct fields names in order to construct a valid solr query.
    Do not explain your approach.
    Always return the response to the question asked by using the result from Kada.
    Do not include citation references in your response. When including hyperlinks, do not display the full URL. Instead, use Markdown link syntax with the word LINK as the display text.

    Click Save

  4. Go to the Knowledge tab. Click the + Add Knowledge button.

  5. In the "Upload files" dialog:

    • Click Upload file or drag and drop the files provided by KADA.

    • Add KADA provided knowledge files:

      • kada_default_columns.txt

      • lineage_rules.txt

      • solr_kada_search_roles.txt

    • Click Add to agent

      image-20251024-023952.png

  6. Go to the Tools tab. Click on the Tool added in Step 5

  7. In the Details section, click on Additional details to expand the options

  8. Set Credentials to use to Maker-provided credentials

    image-20251024-025647.png

  9. Click on Publish to complete the setup of your agent

Step 8: Testing your agent

Test your agent to see how it responds to the following questions

Question

Expected response

What report should I use for <use case>

Agent returns accurate list of reports suitable for the use case

Who is the owner of this report?

Agent returns the correct owner for the report

If the agent provides inaccurate responses, consider tuning the instructions.

Please advise the Kada account manager so your feedback can be incorporated in the MCP / Knowledge files.

Step 9: Deploying your agent to Teams

  1. Navigate to your agent from Agents in the navigation pane.

  2. Go to the Channels tab. Click on Teams & Microsoft 365 Copilot

    image-20251024-024605.png


  3. Select the checkbox if you want to make the agent available in Microsoft 365 Copilot.

    Click Add Channel

    image-20251024-025908.png
  4. Click on Edit Details in the Agent Preview section

  5. Add an icon, change the colour and update the description. Click Save

image-20251105-025335.png

Step 10: Approving the agent deployment in Teams

Requires Microsoft Admin role

  1. Go to https://admin.teams.microsoft.com/

  2. Go to Teams apps in the left navigation pane

  3. In the apps list. Click on the Filter icon. Apply the following filter

    image-20251105-025824.png

  4. Click on your app. Click Publish

Step 10: Adding your agent in Teams

  1. Open your Microsoft Teams application

  2. Click on Apps in the left navigation pane

  3. Click on the Agent you deployed. It will be located in the Built for your organisation section

    image-20251105-025433.png


Troubleshooting

Connection Issues

If you see an error message saying "Let's get you connected first, and then I can find that info for you. Open connection manager to verify your credentials. Once the connection is ready, retry your request."

Fix the connection:

  1. Click Open connection manager (or go to Connections via the left navigation)

  1. Find "Test server KADA" in the connections list

  2. Status shows "Not Connected"

  3. Click Connect button

  1. Pick The KADA MCP connection.

  2. Click Submit

  3. Once Status change to Connected, go back to your agent chat and try again.

JavaScript errors detected

Please note, these errors can depend on your browser setup.

If this problem persists, please contact our support.