Claude Code integration

Claude Code, connected to your data stack. In 1 day.

i-spark configures Claude Code into your dbt and Power BI environment, governed from the start, tested against your actual setup, and handed over so your whole team can use it from day one.

For those that don’t have Claude Code connected to their data stack, i-spark configures it properly, governed from the start, and hands it over in one day.

Why do you need a proper Claude Code setup?

Getting it properly connected is important.

Without a proper setup

Claude Code guesses at your model. 

And here's the thing, it guesses pretty well. The output looks completely fine. It looks like something a competent engineer wrote on a Tuesday afternoon. So you take it, you try to use it, and then it starts breaking. Table names that don't exist in your project. A date table structure that doesn't match yours.

And you're sitting there thinking, okay, who has to fix this? It's the one person on the team who knows the codebase well enough to spot what's wrong, the same person who could have written it correctly in the first place.

With a proper setup

Claude Code reads your project before it generates anything.

It knows your tables, your sources, the relationships you've defined, and the tests you've already written. So when you ask it to generate a new staging model, it's looking at what you've already built and working within that.  Same with DAX. A measure for revenue year-over-year will reference your actual date table, your actual sales table, the relationship between them as it exists in your model. The output is specific.

And when something is specific, people across the team can read it, verify it, and use it.

Configured for your team

Tested against your actual models

Not a generic install. The configuration is validated against your project before handover. You don’t inherit a setup that works in theory.

Governed from the start

Credentials stored securely, rotation documented, ownership assigned. The Power BI connection is more credentials-sensitive than most teams expect, we handle this at setup, not as an afterthought.

Documented so everyone can use it

The MCP server configuration lives in the repository alongside everything else. The next person who joins knows how to use it. No tribal knowledge required.

Full handover, no dependency

You end the day with a walkthrough. The setup is yours, there is no ongoing dependency on i-spark. You own it completely.

Get everything you need to set up Claude with Power BI, in one guide.

Once it's properly wired in,
the whole team benefits

The connection changes what’s possible for everyone on the team, 

not just the person who built the setup. Here’s what it unlocks, per product.

  • 01. Write DAX that fits your actual model. Anyone on the team can describe what they need in plain language and get a working first draft back — DAX that references your actual tables, your actual relationships, your actual date table.
  • 02. Understand measures nobody remembers building. Complex nested calculations written by someone who left, no comments, no documentation. Claude Code reads the expression and explains what it does, step by step.
  • 03. Write Power Query M without memorising the syntax. Describe the transformation in plain language. Get working M code back — null handling, edge cases and all.
  • 04. Plan data models from reporting requirements. Describe what you need — dimensions, aggregations, time intelligence. Get a suggested table structure and relationship design before you build.
  • 05. Document your dataset systematically. Claude Code reads the DAX expressions across your dataset and generates descriptions for measures and columns — covering in hours what would otherwise sit undone indefinitely.
  • 01. Build new models in minutes, not hours. Describe a new source once. Get a working draft of the SQL, YAML, tests, and documentation.
  • 02. Refactor accumulated technical debt. Claude Code reads an existing model and restructures it into clean, modular dbt — CTEs split properly, naming conventions applied, logic that should be a ref flagged as such.
  • 03. Debug failing runs with context. Describe the error in plain language. The diagnosis accounts for your actual model structure — not a generic error lookup.
  • 04. Catch up documentation at scale. Undocumented dbt models are one of the most common pain points in data teams. Claude Code works through existing models systematically — reading what’s there and generating descriptions across the project.
  • 05. Answer questions about your data model in plain language. Which models depend on this source? What tests exist on this column? Where is this metric calculated? No grep, no lineage graphs.

How to connect Claude Code to dbt and what a proper set-up looks like

What WE bring

The setup itself is not complex.

The part that takes care is making it production-ready, tested, governed, and built so the whole team can use it rather than just the person who configured it. That is what the sprint delivers.

Configuration that lives in the repository

 

The configuration is set up to live in the project repository alongside everything else, so every member of the team has access to the same setup from day one.

Governance agreed and documented

 

What Claude Code is expected to do, what it is not, and what the review process looks like, are agreed upfront and documented, so the boundaries are clear before anyone pushes AI-generated code to main.

Tested against your models before handover

 

The setup is verified against your real project structure, not a demo environment. You see it working on your data before the day ends.

Right-sized for your environment

 

dbt Core and dbt Cloud have different connection patterns. Power BI workspaces vary in what they expose. The configuration is matched to your specific version and setup.

Build for how you work

 

The layout adapts to your current state at the office, on the go or the big screen during important C-level meetings.

You get a complete mockup before we build

 

Before you commit to the final look of a dashboard, we offer a complete mockup, you can share your feedback and ensure the final result meets your expectations.

Book the intake call. We'll confirm fit before anything else.

Explore our other data products

Dashboards, assessments, stack setup, and custom builds. Backed by expert analysis to support data-driven decision-making and strategic growth.

Power BI dashboard

Custom Power BI dashboards on top of your existing data sources, connected to your live systems, designed around your business logic, and built so you can open them and trust what you see immediately.

Looker dashboard

Looker implementation from the ground up: LookML data model, custom dashboards, and access governance for  businesses that need a governed analytics layer they can actually rely on.

Looker Studio dashboard

Fully custom Looker Studio dashboards on top of your Google data sources: GA4, Google Ads, Search Console, BigQuery, so you gain a live, accurate view of performance.

Klipfolio dashboard

As an official Klipfolio partner, we build custom dashboards for real-time KPI monitoring across any data source, deployed faster than heavier BI platforms and backed by certified partner expertise.

Simplicate dashboard

If your business runs on Simplicate, we connect directly to your environment and build a custom dashboard that gives you a live overview of hours, projects, pipeline, and margin.

Data Assessments

Before you invest in data tooling, platforms, or hires, a data assessment gives you an honest picture of where your setup stands today: what is working, where the gaps are, and a plan for what to do next.

Power BI/dbt x Claude Code setup

Connect Claude Code directly into your dbt and/or Power BI environment in 1 day, tested against your models and datasets; handed over with a walkthrough so you can use it immediately.

Custom B2B data solution

For B2B agencies and consultancies that need practical data insight, we build custom solutions covering live dashboards, structured data outputs, automated KPI alerts, and interactive reports.

Fully custom data product

Looking for something you don’t see in our modular product range? Our data and AI expertise goes well beyond what is listed here. Tell us what you need and we will build it.