MCPs: What they are and how you can use them with Hex

We sat down with Fran Orozco, a data leader at BloomGrowth, and learned how he’s taking advantage of MCPs to streamline his data dev process. It was pretty sweet so we asked him if he’d be willing to share — and to our delight, he did!

Before we jump in…. a quick MCP intro

You've probably seen MCPs generating buzz on X. But what are they really? 🤔🤔🤔

Model Context Protocols create a standardized API for AI systems to access context and use tools. MCPs define how:

1. Tools expose capabilities to AI assistants — and you can define those tools

2. AI models request actions using those tools — and then can combine multiple tool use

3. AI models can automatically write code to quickly parse/extract the necessary info returned from the tool calls

It's really easy to plug in APIs that already exist, customize them to your use case, and get started with tools like Cursor or Claude.

So, what does this look like in practice?

That’s where Fran comes in! He’s been experimenting with MCPs and experiencing first-hand how they're saving him time. Let’s turn it over to Fran 🙂...

MCPs: Rethinking the dev experience

👋 Hi — I'm Fran, a data lead at BloomGrowth! We're a team that’s quickly growing, and so is our use of data to ask and answer questions. This means Hex is an integral part of our data stack and is used throughout the business.

While Hex provides an incredible frontend experience for ad hoc analysis and interactive reports, a lot of my day-to-day analytics engineering work is done in Cursor. I'm constantly doing data transformation using dbt, building tests, or writing documentation. After this, I need to re-execute my Hex projects to make sure they've run successfully, but I've always wished I could do this from my own development environment without switching tools and contexts.

This workflow is changing dramatically with the introduction of MCPs. By using MCPs in Claude or Cursor, our developers and data engineers are accelerating development times, simplifying onboarding challenges for new engineers, and orchestrating jobs using only natural language.

So how does this actually work day-to-day?

I have a Hex project that pushes data to HubSpot. I run it daily, but when I integrate new data, I want to run the project ahead of the usual schedule. I can ask Cursor to search for it:

It returns the project along with metadata like its schedule. I can ask for it to run it again and give me a run status:

Finally, I can ask if the project has finished running:

This unlocks a new level of interactivity as a data practitioner working on the backend and makes Hex feel integrated across the dev data stack. This new workflow isn't just about convenience — it's about maintaining deeper focus. My team can iterate faster on changes and observability is a breeze.

Under the hood: How my Hex MCP server works

My open-source hex-mcp project bridges Hex's existing public API endpoints with the MCP standard.

Today, Hex offers APIs that allow you to:

• List and review projects

• Trigger project runs

• Check run statuses

• Cancel ongoing runs

The hex-mcp package wraps these capabilities in the MCP format, enabling AI assistants like those in Cursor or Claude to understand and use them. It essentially translates between:

AI Assistant (using MCP) → hex-mcp server → Hex API → Hex Platform

This translation layer means I can interact with Hex through natural language in my dev environment and the MCP server handles the details of API authentication, making requests, and response parsing.

Set up your own MCP integration: A step-by-step guide

Want to give my MCP package a spin? Follow along below! My MCP server is open-sourced, so also I welcome folks to contribute or fork their own versions specific to their needs 🙂.

1. Install the hex-mcp Package

Using uv (recommended for a cleaner environment):

bash
Copy
uv add hex-mcp

Or with traditional pip:

pip install hex-mcp

Verify the installation:

hex-mcp --version
# Should show: hex-mcp version 0.1.10 (or newer)

2. Configure your workspace key

Set up your Hex API key:

hex-mcp config --api-key YOUR_HEX_API_KEY
# Configuration saved to ~/.hex-mcp/config.yaml

This saves your API key in your home directory, making it accessible across projects without environment variable hassles.

3. Install MCP in your project

In your project directory, run:

hex-mcp install
# MCP configuration installed to .cursor/mcp.json

This creates the necessary configuration file that Cursor uses to discover and interact with the MCP server.

4. Start using it in your dev environment

Open your project in Cursor, activate the AI pane, and start asking questions about your Hex projects! Try prompts like:

• "Find my HubSpot data integration project."

• "Run the weekly sales dashboard project."

• "Is my customer churn analysis project done running?"

The future of observability and governance

This integration between Hex and Cursor through MCPs represents just the beginning of a broader shift in how data teams can take advantage of AI. My team no longer has to jump from one context to another — our development workflow is more streamlined and our interaction with Hex as a platform is much more enjoyable.

With expanding product API access (spoiler: Hex is working on this!) — you can imagine observability and governance to be far easier. Imagine a world, where, with just natural language you could:

• Handle project lifecycle management at scale — automatically archiving or deleting projects based on specified rules

• Do bulk handling of user and group permissions

• Create and modify data connections

This list goes on! By using MCPs, now much of the painful and tedious administrative tasks can be handled with a prompt — lowering the barrier to entry and improving Hex usability for everyone on our team.

Want to contribute or explore more? Check out my GitHub repository for the Hex MCP implementation or dive into the Model Context Protocol specification to learn how you can customize an MCP for your team!

PSA from Olivia: At Hex we're excited to expand our public API coverage — making it easier than ever to manage your Hex instance at scale (and especially with the help of LLMs!). We’re looking for design partners for these changes so if you're interested in providing API feedback for better workspace management please reach out at magic [at] hex.tech.