A hosted Model Context Protocol server that puts the Austin MLS inside Claude, ChatGPT, Perplexity, Cursor, and Gemini CLI. No install, no repo to clone, no API key to generate — paste one URL into your MCP client and you’re searching live listings.

This is the MLS-specific, hosted sibling to the Local Austin MCP. That one’s open-source and free, covers the whole civic-data surface of Austin, and includes basic listing search. This one is MLS-specific, hosted, and gated by tier — free for anyone, deeper (sold comps, market stats, neighborhood stats) for clients with a signed buyer representation agreement. Two different tools, two different jobs — don’t confuse them by name.

What it does

Eleven tools covering the two things people actually ask an AI about real estate: what’s for sale, and what’s it worth.

Listings — search by city, ZIP, district, price, beds, baths, and property type. Pull full detail on a specific MLS number. See the newest listings in an area. Free tier sees active and active-under-contract only, last 30 days. Active Buyer tier sees every status, full listing history.

Market intelligence — sold comparable sales, median-price and inventory statistics by area, neighborhood-level stats. Gated to Active Buyer tier — this is the data that actually helps someone decide what to offer, so it’s reserved for clients who’ve committed to working with the brokerage.

Content — search the Neuhaus Realty Group blog and Complete Guides, pull the latest articles. Free on both tiers.

How it’s built

Node.js on the MCP Streamable HTTP transport, running on Neuhaus Realty Group’s VPS behind Cloudflare and nginx. It shares the same Postgres database and Redis cache as the VOW property search that powers neuhausre.com’s site search — one dataset, two front doors.

Connection uses full OAuth 2.1 with dynamic client registration and PKCE, the same handshake every major AI client uses for a custom connector. First-time users get a one-click “continue free” flow that captures name, email, and phone, provisions a rate-limited token, and — this is the part I like — automatically creates a Follow Up Boss contact tagged Claude MCP. Someone asking their AI about Lake Travis listings becomes a lead the same way someone filling out a contact form does. No manual entry, no gap.

Every listing link in a response carries UTM tracking back to the source. Every search from a client we already know pushes a timeline note to their CRM record, so if someone’s been asking Claude about 3-bedroom homes in 78738 for a week, that shows up before the follow-up call.

Rate limiting runs at both nginx and Redis — sliding-window quotas per client, per hour and per month, so free-tier usage stays free-tier cheap while paid clients get real headroom.

Why I built it

The Local Austin MCP proved people will connect an AI assistant to real Austin data if the setup is trivial. But that project is deliberately keyless and public — it can’t gate anything or know who’s asking.

This one can. It’s the version built for the actual business: free enough that anyone curious about Austin real estate gets real value from it, structured so that the moment someone’s serious — signs a buyer rep agreement — the same chat interface unlocks the comps and market data that actually inform an offer. The AI conversation becomes the intake form.

Connect

https://mls.neuhausre.com/mcp

Paste that as a custom connector in Claude Desktop, ChatGPT, Perplexity, Cursor, or Gemini CLI. Full pricing, setup walkthroughs per client, and use-case examples live at neuhausre.com/austin-mls-mcp. The GitHub repo is intentionally a pointer, not the source — this service runs on Neuhaus Realty Group infrastructure and isn’t distributed for self-hosting.

License

All rights reserved. The connector repo ships no server code — connect to the hosted service instead.