“Interacting naturally with the full breadth of MLS data elevates both the level of insight and the confidence behind it. For subscribers, that shows up in better decisions, faster workflows, and more informed client conversations. For MLS organizations, it reinforces what they already are: the source of truth for the market.”
— Kim Prior, VP of Product, FBS
OK, so here’s what FBS actually built: a way for Flexmls subscribers to connect ChatGPT, Claude, or Gemini directly to their MLS account using the Model Context Protocol (MCP). You ask a question in plain English, and the AI answers it using your MLS’s actual, complete, standardized listing database.
The nerdy detail that matters here is MCP. It’s an open standard that’s quickly becoming the way AI tools connect to outside data sources. FBS building on it (rather than rolling their own proprietary thing) is the right call. It means subscribers can move between AI platforms freely. Use Claude today, switch to Gemini tomorrow. The plumbing stays the same.
And this is where it gets interesting for the MLSs themselves. The Flexmls MCP Server works for agents pulling comps, sure, but the MLS can use it too. Market reports, local intelligence, all generated from their own data. That’s the MLS reinforcing its position as the authoritative source in its market. Which, honestly, is something a lot of MLSs need to be doing more of right now.
Each MLS controls activation. Authentication runs through existing Flexmls credentials. The data governance stays where it should: with the MLS. FBS clearly thought about the security question before anyone could ask it.
What I’m curious about is adoption. Sure it’s seems cool, but will MLS members even know what to ask? Is a chatbot the way forward? Plus there are real estate agents, and AI agents. How do we account for those use cases?
That being said, 330,000+ subscribers across North America now have a path to plug AI into their actual workflow. Real connection to real data, ready to go.
Good explanation, Greg, thanks for the post! You’re right that most agents will just want the power right within the MLS, which we’ll be launching soon as well.
Greg – I would love to see you and dan collaborate to write a prompt against this solution to build a CMA that challenges CloudCMA and contrast the meaningful differences.
I think there’s potential to generate a few one-off CMAs using prompts with the right model and data. But Cloud CMA is still the industry’s best CMA tool because it provides a simple, consistent workflow for creating personalized, custom reports, without the risk of hallucinations.
There are still plenty of W+R folks at Lone Wolf that I care about, and I know they’re actively working on leveraging AI to make Cloud CMA even better.
I think you’re right to be concerned about adoption. We were talking to boards about offering AI features a few months back and an MCP server was one of the options – there were a lot of blank stares amongst the staff. Even using tools like Gemini in Chrome that are widely available that can be used to augment research, etc. didn’t seem to be that widely used. I’m sure power users are going to have fun with this but it feels like if there isn’t a top-down effort by boards to explain & train all this to their members there may not be a lot of adoption.