I've been in data long enough to remember when "self-service analytics" meant you could run your own queries in SQL Management Studio instead of asking someone else to do it for you, but this is actually different. The Qlik MCP (Model Context Protocol) server with Claude is the kind of thing that makes you sit back and think "Okay, maybe we are living in the future". Then you immediately try to ask it something and get confused about OAuth scopes, and you're firmly back in the present.
Don't worry. I've done the confused part so you don't have to. Let's dig in.
First, What Even Is MCP?
MCP stands for Model Context Protocol. It's an open standard created by Anthropic that lets AI assistants like Claude securely connect to external tools, data sources, and services. Think of it as a universal adapter the kind that actually works, unlike those cheap travel adapters that melt your laptop charger.
In plain terms: MCP is what allows Claude to start actually talking to your systems. Instead of you copying data out of Qlik and pasting it into an AI chat window (we've all done it, no judgment), Claude can go directly to the source.
Qlik launched its official MCP server to in February 2026, enabling third-party assistants including Anthropic Claude to securely access Qlik's analytical capabilities and trusted data products. So this isn't a hacky workaround it's built for production.
The Architecture (Without the Architecture Diagram Nobody Reads)
Here's what's happening under the hood:
Your Qlik Cloud tenant exposes an MCP endpoint at <your-tenant-url>/api/ai/mcp. Claude.ai (or Claude Desktop) connects to that endpoint via OAuth. Once authenticated, Claude can call a set of tools exposed by the Qlik MCP server things like searching for apps, querying data, making selections, building charts, managing spaces, and a whole lot more.
The key thing Qlik is emphasising here is governance. An assistant can request analysis based on Qlik-managed measures rather than generating an answer from copied data or a user-provided spreadsheet.
In other words, the numbers Claude gives you are your numbers, calculated by Qlik, using your business definitions. Not hallucinated. Revolutionary concept, I know.
Setting Up the Qlik MCP Server with Claude.ai
Right, let's get our hands dirty. I have only tried the official Qlik Cloud way, although if you are brave enough you can go down the self-hosted open-source way (more control, more configuration, more opportunities to question your life choices).
Step 1: Enable AI Features on Your Tenant
Before anything else, your Qlik Cloud tenant needs to have AI features enabled and you will need to enable cross-region data processing. this can be found in you administration settings at the bottom of the page.
Step 2: Allow users access to the Qlik MCP server
This is done by user role permissions, setup a new role if one has not already been created and under Features and actions > Agentic AI set MCP to Allowed.
Users can then be granted this role giving them access to Qlik MCP server
There's also a default shared Client ID (76d3f46e87655a50424bec7e0f0bb1e2) that Qlik provides for Claude specifically. Depending on your tenant's security setup, you can use either this or your own custom one.
Step 3: Connect Claude.ai to the Qlik MCP Server
Now the fun part. In Claude.ai:
- Open Settings
- Navigate to Connectors
- Click Add custom connector
- Give it a name - something professional like "Qlik MCP"
- For the Remote MCP Server URL, enter:
<your-tenant-url>/api/ai/mcp - Under Advanced settings, enter your OAuth Client ID
here is the default shared Client ID (
76d3f46e87655a50424bec7e0f0bb1e2) that Qlik provides for Claude specifically. - Leave the OAuth Client Secret field empty
- Click Connect
- Sign into your Qlik Cloud tenant when prompted
- Click Approve
- In your Setting > Connectors you will now see your newly created custom connector, click on Configure and set all Tool permissions to Always allow
And that's it. You're connected. Claude can now see your Qlik world.
So What Can You Actually DO With It?
Now for the part everyone actually cares about. Here's just some of the things that I have played around with while testing.
Discover and Navigate Your Qlik Environment
You can ask Claude to list and navigate everything in your Qlik Cloud tenant. Apps, spaces, sheets, datasets all of it becomes conversational.
“Show me all the apps in the Sales space”
“What sheets does the Revenue Dashboard have?”
”Find any app related to supply chain”
This sounds simple but it's genuinely powerful when you're working in a large tenant, instead of looking for that one app someone built two years ago and named something deeply unhelpful, you just ask.
Query and Analyse Data From Your Apps
This is the headline feature. Claude can reach into your Qlik apps and actually compute against your data using the Qlik associative engine.
“What were total sales by region for Q4 2025?”
”Show me the top 10 customers by revenue”
”What's our average order value this month vs last month?”
The critical thing here and it's worth saying again is that these calculations are done by Qlik's engine, not guessed by Claude. The associative model, your master measures, your business logic all of it stays intact. Claude is the natural language interface; Qlik is doing the calculations.
Apply Filters and Selections Conversationally
One of my favourite features. You can apply selections to your data through conversation, exactly as you would clicking in a Qlik sheet.
“Filter to only show data from the UK”
”Select Q1 and Q2 only”
“Clear all selections and start fresh”
You can even use advanced patterns like range queries: "Show me only customers with revenue greater than £100,000". Claude translates that into the right Qlik selection syntax so you don't have to think about set analysis expressions. (Though if you're a Qlik nerd who enjoys set analysis, nothing is stopping you from writing your own.)
Create Charts and Visualisations
Yes, Claude can actually build charts inside your Qlik apps. Not mock-ups. Not suggestions. Actual charts on actual sheets.
“Add a bar chart to my Sales Overview sheet showing revenue by product category”
”Create a KPI showing total headcount”
”Build a line chart of monthly revenue with a moving average”
It supports bar charts, line charts, pie charts, scatter plots, KPIs, tables, treemaps, heatmaps, combo charts, and pivot tables.
What to Watch Out For
A few things worth knowing before you start using this in anger:
Capacity limits are real. A tenant's Qlik MCP server can have a maximum of 20 concurrent application sessions loaded into memory at one time, and there is a maximum of 300 MCP tool calls per user per minute through the Qlik API gateway. For most use cases this is fine. For power users hammering complex queries all day, plan accordingly.
Sessions time out. OAuth sessions don't last forever. If your session is disconnected, access the LLM client's settings to establish a new session. This is slightly annoying but takes 30 seconds to fix.
Governance conversations still need to happen. The data is governed by Qlik, but you're still sending queries and results through Claude. Make sure your data governance and security teams know what data you're putting into this workflow especially for anything regulated or sensitive.
Monthly question quotas. Depending on your subscription, you have a monthly quota of questions. Questions do not carry over between months. When you hit your limit, users can no longer ask questions until the next month.
Quick Reference: Useful Things to Ask Claude Once You're Connected
Here's a cheat sheet of prompts to get you going:
"List all Qlik apps I have access to"
"What sheets are in the [App Name] app?"
"Show me total revenue by country for 2025"
"Filter to just Q4 and show me the top 5 products"
"Clear all my current selections"
"Add a bar chart to [Sheet Name] showing [measure] by [dimension]"
"Search for any dataset related to customers"
"Show me the lineage for [Dataset Name]"
"Find the glossary definition for [Term]"
"What's in the [Space Name] space?"
Wrapping Up
The Qlik MCP server with Claude.ai is genuinely one of the more exciting developments in the BI space in a while and I don't say that lightly, given that I have to get excited about data for a living.
The setup, as we've covered, isn't particularly painful. The native Claude.ai integration is a few clicks and an OAuth dance. The open-source path is a bit more involved but gives you flexibility. Either way, once you're connected, you've got a surprisingly capable interface to one of the most mature analytics platforms out there.
And if it turns out that your executives really do start just asking Claude questions about your data instead of scheduling a meeting to ask you to make a chart well, look on the bright side. You can spend that freed-up time building something even better. Or finally understanding what your own set analysis expressions actually do.
Either way, have fun with it. This stuff is cool.
Have questions about the setup? Found something broken? Discovered a use case I haven't mentioned? Drop a comment below or, fittingly, just ask Claude about it.
Top comments (0)