[Video] Qlik MCP Server is Here: Enable & Configure it with Claude

[marksouzacosta]

Hi Qlik Community,

The new Model Context Protocol (MCP) Server is now available in Qlik Cloud! This is a fantastic new capability that allows us to standardize how we connect AI models to our Qlik data.

To help everyone get started, I've recorded a complete, step-by-step video tutorial on how to enable, configure, and connect the Qlik MCP Server to Claude Desktop.

📺 Watch the video tutorial here: https://youtu.be/eUo0y8j3ITU

The video covers:

1. Enabling the AI features in the Qlik Cloud Tenant.

2. Creating a Qlik MCP permission and assign users.

3. Adding Qlik MCP Connection to Claude Desktop

4. Authenticating and testing the Claude Connection to Qlik MCP Server

For a written companion to the video, you can also check out the full article on the Data Voyagers blog here: 
https://www.datavoyagers.net/post/qlik-mcp-server-is-here-enable-configure-it-with-claude

I'm excited to see what use cases you all come up with for this integration. Let me know your thoughts in the comments!

Regards,

Mark Costa

[F_B]

Thank you for sharing.

[igoralcantara]

This is a very good video, @marksouzacosta. Thanks for sharing.

[AnaSilva]

Thanks for sharing this, @marksouzacosta ! That's very helpful!

[hugo_andrade]

Thanks Costa! Really nice!

[priscilarubim]

Thanks for sharing.

[Antony3]

Hi marksouzacosta, Thank you for the video. I have been asked to look at integrating Claude with Qlik for some users within my org and I have few questions. Hopefully its ok to ask.

Questions
- How does usage/tokens work? I notice that in my admin panel that Qlik has a limit of 200 for "Assistant questions asked (monthly)". If someone prompts 200 times in Claude does it use this limit and is 1 prompt equal to 1 question or are are questions unrelated to external tools? How would I track prompting at tenant level if possible?
- If we wanted to give specific users write access is there a way to limit their ability to do things that may impact our pipeline. For context we don't give the users requesting write access analyser level licenses and limit their ability to manage create things in Qlik itself. If an existing user doesn't have access to a specific managed space does that mean Claude also wont have access to that space? I could in theory block all Qlik users from the development apps space so its just me. 

- It seems like the read/write tool settings are done in individual instances of Claude rather than Qlik side? Its unlikely, understandably, that senior management would let me log into their Claude instance to set up their connection. Can it be done at a Claude business level if that's a thing? i.e. Our head of IT creates a connection/role in Claude and manages the permissions at that connection.

- If I wanted to set myself up to do some testing how could I ensure that I only work in a designated space when prompting? i.e. I create a copy of a development apps in a Claude Apps managed space and then only prompt from Claude into that space? 

Context if helpful
I currently maintain the entire tenant solo, including coding, visualisations and automations. Some senior management have been using Claude in excel and they were looking to do the same with Qlik. Our data pipeline works as follows; extract from sources, transform1 e.g. customers table, contracts table etc, transform2 (app specific transformation), live apps, automations. I am partially worried about giving people write level access and then finding things have been changed and or deleted based on prompting. It works for us in excel because its usually slabs of exported stuff but a misjudged prompt might torpedo all of our reporting. 

[marksouzacosta]

Hi @Antony3, thank you for the questions.

1. How does usage/tokens work? I notice that in my admin panel that Qlik has a limit of 200 for "Assistant questions asked (monthly)". If someone prompts 200 times in Claude does it use this limit and is 1 prompt equal to 1 question or are are questions unrelated to external tools? How would I track prompting at tenant level if possible?

When using the Qlik MCP Server on external LLM software like Claude and ChatGPT, 5 Tools executions consume 1 "Assistant questions asked". From Qlik Cloud Subscription Documentation - https://assets.qlik.com/image/upload/v1713309774/qlik/docs/Legal/terms-and-conditions/product-descri...

Example: if you get 5 Dataset Samples in your Claude session using Qlik MCP Server, it will call the qlik_get_dataset_sample tool 5 times, consuming then 1 Question.

2. If we wanted to give specific users write access is there a way to limit their ability to do things that may impact our pipeline. For context we don't give the users requesting write access analyser level licenses and limit their ability to manage create things in Qlik itself. If an existing user doesn't have access to a specific managed space does that mean Claude also wont have access to that space? I could in theory block all Qlik users from the development apps space so its just me. 

Qlik MCP Server reflects exactly the very same security set on the Qlik Cloud side. So, if a user don't have access to a Space in Qlik Cloud, it won't have access to that Space using Qlik MCP Server or any other Qlik official tool - like Qlik-CLI, Qlik Rest APIs, etc. It is all integrated - including Section Access.

3. It seems like the read/write tool settings are done in individual instances of Claude rather than Qlik side? Its unlikely, understandably, that senior management would let me log into their Claude instance to set up their connection. Can it be done at a Claude business level if that's a thing? i.e. Our head of IT creates a connection/role in Claude and manages the permissions at that connection.

That is correct, you can't control the availability of MCP Tools at an Enterprise level in Claude - I guess they may release that at some point. Is up to the Claude user to allow a MCP Tool to execute or not at his chat level.

4. If I wanted to set myself up to do some testing how could I ensure that I only work in a designated space when prompting? i.e. I create a copy of a development apps in a Claude Apps managed space and then only prompt from Claude into that space? 

That works. You can even create a copy of any app and use that app ID to make sure you will not be taking any other app accidentally. Always try to use IDs instead of names when using the MCP Server. This is usually the most secure way, and it is also faster to obtain answers.

Regards,

Mark Costa

[Sunilsahu_345]

Thank you