Of course, we built an MCP server

As MCP servers are everywhere these days, we knew we had to explore what this technology could unlock for Validio. An initial prototype in Python quickly showed that this might be an extremely helpful way to multiply the insights our platform delivers. In this post, we will walk you through what an MCP server is, why we think it’s an excellent addition to the Validio developer tools, and go into detail on some of the use cases we find most interesting.

🚩 TABLE OF CONTENTS

→ What is MCP (and why should you care)?

→ Why a Validio MCP server makes perfect sense

→ Use case 1: Root cause analysis with lineage

→ Use case 2: Fast monitoring setup with data profiling

→ Use case 3: Fix a data pipeline and create a GitHub PR

→ Why does the Validio MCP server work so well?

What is MCP (and why should you care)?

MCP is an open protocol that solves the "N×M integration problem". Instead of building custom integrations between every AI tool and every data source, you implement the standard once and gain compatibility with any MCP-enabled application. In other words, your AI Assistant (Claude, I choose you!) can connect to an MCP server like the GitHub MCP servers and fetch data from your repos, analyse code or merge pull requests. Now this is great on its own, but the AI Assistant can obviously connect to multiple MCPs at the same time, moving data and insights between services intelligently. 

“Hey Claude, check if the latest incident in Validio was caused by an update to the data pipeline code” -> the AI Assistant will now query both Github MCP and Validio MCP and do its analysis. Hands off the keyboard for you. 

The architecture that enables this is pretty simple: you need an MCP server that hosts the Tools (executable functions), Resources (structured data for instruction), and Prompts (reusable templates) and an MCP client (Cursor, Claude Code, Gemini-cli, etc).  The MCP server for Validio is hosted by us, so all you need to do is connect to it with your favourite client. 

Since Anthropic introduced MCP in November 2024, the ecosystem has exploded. GitHub, Stripe, Sentry, Atlassian, and dozens of others have built MCP servers. OpenAI, Google, and Microsoft have all adopted the protocol, and it's becoming the standard way AI assistants interact with external systems. For Validio users, this means more and more ways to combine data quality insights from Validio with your other systems in the organisation. 

Why a Validio MCP server makes perfect sense

Even if you don't connect other tools to Validio in your AI chat, the Validio MCP offers significant value for data practitioners on its own. As an analyst or data engineer relying on high-quality data and trustworthy metrics, you need to investigate data quality issues quickly and thoroughly.

Your daily routine involves monitoring freshness, tracking validation failures, investigating anomalies, and coordinating fixes across teams. Most of this work centres around questions like:

• "Which pipelines failed in the past 24 hours?"
• "What's causing the spike in null values for the users table?"
• "Is the customer_orders source still fresh?"
• "Which upstream incident is affecting my dashboard?"

Answering these questions often requires jumping between Validio dashboards, data warehouses, documentation, and messaging tools. What if you could answer all these questions in one place, using natural language? That's exactly what the Validio MCP server enables. Below, we show you how this works in practice:

Use cases:

Let's walk through what becomes possible when Claude (or any MCP-enabled AI tool) can talk to Validio.

Use case 1: Root cause analysis with lineage

Data quality incidents often propagate through your pipelines. When upstream data arrives late or breaks, downstream Validators start throwing incidents. A large outlier in a Bronze layer table can trigger incidents across multiple tables downstream. Identifying the root cause means understanding how your data flows, which is where Validio's automated lineage becomes essential. Our MCP server can interact with incidents and lineage to automate root cause analysis (RCA) for you. 

Without MCP:
You receive an incident alert in Slack. You click on the link to navigate to the Incidents page, click through to incident details, switch to the Root Cause Analysis tab to identify upstream dependencies, and cross-reference with pipeline documentation and code to determine the root cause.

With MCP:
You receive an incident alert in Slack, so you open Claude and ask: "What happened October 10th on gold sales summary table? Give me the root cause"

Use case 2: Fast monitoring setup with data profiling

Setting up comprehensive monitoring for a new data source requires understanding its data types, data distributions, and potential failure modes. You often need to profile the data, identify critical fields, and configure validators with appropriate thresholds and windows. A process that typically involves sampling queries, statistical analysis, and sometimes a lot of manual configuration.

Without MCP:

You manually query the warehouse to understand the schema of the newly added table, write SQL to sample and profile the data, analyze patterns and statistics, then switch to Validio's UI to create Validators based on your insights. 

With MCP:

You open Claude and ask: “delivery_routing is a new table. Profile the data of the table and recommend key Validators to detect inconsistencies in the data going forward.”

Use Case 3: Fix a data pipeline and create a GitHub PR

When incidents point to pipeline failures, the investigation often extends beyond Validio into your code repositories. You need to trace the issue from the data quality incident back to the transformation logic, identify the bug, fix it, and deploy the change. A workflow that usually spans multiple tools and contexts.

Without MCP:

You check Validio for the latest incidents, check dbt model runs in dbt, open your IDE to find the failing model, identify and fix the bug, create a branch, commit, push, and manually create a PR. 

With MCP:
You ask Claude: "Check for incidents in Validio" Claude detects the Freshness issue and pipeline failure in Validio, finds the recent column typo bug in the dbt code, fixes it, and opens a GitHub PR with full context.

Why does the Validio MCP server work so well? 

Let's be clear: the reason the Validio MCP server works so well is not because LLMs are magic. If an AI assistant already magically knew everything about your data, how to monitor it, and how to fix issues, data quality would be a solved problem, and you wouldn't be reading this (If you're a human, that is). The reason the above and countless other use cases work so well is that we have engineered the MCP server to have the right tools in place for any LLM to leverage. We provide specific routes and tools that make it easy for the LLM to fulfil your tasks. These tools enable listing and filtering catalogue assets, as well as inspecting incidents. They also combine Validio's complete internal state, from lineage and catalogue to data quality, into easily parsable data for an AI assistant.

All our purpose-built tools rely on our optimised internal APIs and allow for fast and consistent execution of actions in the platform, as you are used to when using our GUI.
We have further added detailed instructions on how Validio works, how to set up Validators, and how to achieve a comprehensive setup at scale.  This includes our own best practices for data quality monitoring and business metrics monitoring. We encourage users to supplement this with their own business-specific instructions in CLAUDE.md or GEMINI.md style files containing guidance. These instructions can contain your naming conventions, critical data assets, root cause analysis procedures, and business metrics monitoring priorities. The AI assistant reads this context and tailors its recommendations to your environment.

Unlike many MCP servers that often simply expose CRUD operations and leave users disappointed, the Validio MCP delivers genuine productivity gains. Because MCP is an open protocol, you're not locked into a single AI provider. Use Claude, ChatGPT, or any MCP-enabled assistant that fits your workflow. And as more tools adopt MCP, you will unlock even more ways to combine Validio with the rest of your data stack.

Try it yourself - the first MCP you’ll actually use?

If you're already using Validio, give it a try and get started with our user guide here. If you're not using Validio yet, request a demo and see what conversational data quality looks like.