A category just got defined in the Claude directory

Every Tuesday morning I do something I call MCP reconciliation.

I sit down with my coffee, open up Claude, and walk through my connections. I update permissions on the connectors I already use. I scan what’s new in the directory. I think about which ones I want to test for client work, and which ones might change how my studio operates.

Most weeks nothing dramatic happens. Today…quite the opposite.

I was scrolling through the directory and stopped on a name I hadn’t seen before. Verisk Underwriting Intelligence. ISO Loss Cost insights, available conversationally, inside Claude. Four tools. Built by Verisk Analytics.

Two scrolls down, another one. Verisk XactRestore. Eighteen tools. Six prompts. Natural-language estimating powered by Xactware’s pricing data.

Two insurance-specific MCP connectors. Both new. And I mean brand new. Both belong to Verisk, a well-known data analytics company in the insurance industry.

I went looking for the press release. Verisk had announced it on May 5. Anthropic mentioned it in their finance and insurance agents launch the same day.

Nowhere in any of the announcements does anyone say what actually happened.

Verisk shipped the first insurance-specific MCP connectors in Anthropic’s Claude directory. As of this week, they are also the only ones.

How does this get missed? Is it just me who sees this as kind of a big deal?

Right now, the directory has roughly 400 connectors, across 30 categories. There is no dedicated Insurance category…yet. Verisk’s two products are filed under Finance and Trading, alongside Bloomberg, Dun and Bradstreet, Experian, and a stack of equity-research tools. These are all indicators of this being category defining. So why not announce that?

But, this really isn’t about Verisk.

It’s about evolving your business and growth strategy to capitalize on new channels like an MCP.

The understanding of what MCP means in terms of growth.

Let me back up a minute in case you haven’t heard of MCP. Model Context Protocol is a standard for how AI agents connect to a company’s data, tools, and workflows. When a company builds an MCP server, any compliant agent (Claude, ChatGPT as MCP adoption spreads, Cursor, others on the way) can plug into the data and tools that company decides to expose. Access can live behind a paid subscription.

When Anthropic puts a connector in their directory, three things happen at once.

1. Every Claude user on a paid plan gets a one-click way to bring that company’s data into their conversations. That is millions of knowledge workers, today.

2. That company becomes part of the workflow on the desktop where serious knowledge work is starting to live. The data lives inside the thinking surface, I like thinking of as the substrate, which is waaaaay more relevant than existing as a feature in a tab or a portal that a user has to choose to open. And waaaaay more relevant than being just a connector.

3. And, this is important. By doing this, the company starts shaping how every other AI model and agent will eventually access their category. Shipping the connector gets you through the front door and you’re actually setting the access pattern for an entire industry. Don’t you want to be the one who does that?

Now that’s the real story.

This is why I keep telling founders I work with that MCPs deserve their own line item on the strategy. They are a distribution channel, an access pattern, and a positioning decision rolled into one.

Every founder building something defensible really needs to decide how they are going to play in this new marketplace. Being part of the first 300-400 makes you a default. It gets your brand and technology written into workflows before your competitors even know they exist.

Verisk is one of the most established data providers in insurance. ISO is a Verisk business. Xactware is a Verisk business. Most of the regulatory-grade data carriers, MGAs, and adjusters work with every day runs through them in some form.

By being first into Anthropic’s directory with two distinct insurance products, they made their data the default insurance data layer for anyone using Claude as their thinking surface. It is no surprise to me that they are leading especially with Louis DiModugno as the company’s Global Chief Data Officer. When I saw him back in October, I knew he was up to something cool. Go Louis and team! Way to lead.

Now, to add practicality here. Picture a new MGA underwriter starting at a carrier this summer. She’s told to “use AI more in your workflow.” She opens Claude. She sees a Verisk connector. She pulls ISO loss cost trends conversationally before she ever opens the ISO Net portal. That’s how sticky defaults form.

They positioned themselves as the trusted, regulator-grade option in a category Anthropic is actively building toward. By the time the Insurance category earns it’s place in the directory, Verisk is already in it.

Founders need to look at MCP as a growth strategy

MCPs are a distribution and positioning decision before they’re a technical one. That’s how they need to be viewed first and foremost.

The real questions you need to start asking are what slice of our product wants to live inside AI, what data do we expose, what stays behind the paywall, how does our brand show up when someone is doing serious work in a directory we don’t control?

Access is a strategy.

Here’s how I’m thinking about it with the RevUp studio’s current client work as the lens.

1. Decide which surface of your product wants to live in AI. The right answer is usually the part your customers already query daily, or the part where collapsing twelve clicks into one removes real friction. Verisk picked underwriting indications and restoration estimating because those are the highest-frequency, highest-friction surfaces in their footprint. Pick yours with the same discipline. Think of it also as an entry point that leads to cross- and up-selling experiences. Product-led growth.

2. Build the connector to the standard. MCP is platform-neutral by design. The connector you ship to Anthropic should be the same connector you publish to any other agent surface as the protocol spreads. If you’re building this once per platform, stop doing that.

3. Position the connector inside your brand, not as a side project. Verisk’s listing reads like a Verisk product. Branded icon. The listing itself is real estate. It’s a product page for third-party AI agents and future buyers. Build authority in the message. Remember, we not too far off from most interactions being agent-to-agent (A2A), not agent-to-human (A2H). Can I coin that now?!

4. Plan the AI visibility loop around it. A connector in the directory is a passive asset until people know it exists. The marketing system around it is what turns it into pipeline. Founder thought leadership. AEO answers that surface when buyers query “is there a connector for X.” Social posts that explain the workflow. Sales enablement that points at the connector during a buying conversation. That’s the layer. And dare I say a press release that tells the more interesting story that insurance has finally arrived.

The window is open…but like everyone is saying right now, it’s closing.

If you’re a tech founder. A carrier innovation lead. An MGA CEO. The. time. is. right. this. second.

Verisk got there first as the data layer. There are many other layers that don’t have a default right now.

The category is open. The directory has room. This is a new distribution channel and Anthropic doesn’t even have an Insurance category yet, but it will. And whoever ships connectors before that category exists has a real chance to shape what it looks like.

Verisk could have called this a category-defining moment. I think they should have. On the flip side, it’s easy to say that this kind of move doesn’t need a headline because it shows up in market share data eighteen months later. But sometimes I just want to grab a CEO by the shoulders and say, tell people what you actually just did!

What’s your MCP strategy? Please tell me more.