Connecting Claude to your CRM with MCP

Every real estate platform has an API. That’s been true for a decade, and it has changed almost nothing for the average agent, because an API is a promise that somebody could build an integration if they had a reason, a budget and a developer.

MCP is different in a way that takes a minute to appreciate.

What is MCP, in plain terms?

A standard way for AI tools to plug into other software. The project’s own site calls it “a USB-C port for AI applications”: one kind of connector instead of a different cable for every device.

Anthropic released it as an open standard in November 2024, describing the problem it was meant to solve: “Every new data source requires its own custom implementation, making truly connected systems difficult to scale.” Since then it has been picked up well beyond Anthropic’s own products.

For a CRM, an MCP server is a short menu of things an assistant can do in your account, each described in plain language: search contacts, read a timeline, list today’s appointments, add a note. The assistant reads the menu when you ask something and decides which items it needs.

How is MCP different from an API?

With an API, a person writes the integration ahead of time. With MCP, the assistant works it out when you ask.

With a REST API, someone writes code that says: when this happens, call this endpoint, map these fields, handle these errors. That code is a product decision. It exists because a company decided the integration was worth building, which means it exists for the popular combinations and not for yours.

With MCP, there’s no integration to write, because the “integration” is the model reading a menu and choosing. That sounds like a small distinction. In practice it means the long tail suddenly works. Nobody at our company had to anticipate that you’d want to ask which contacts from a specific source haven’t been touched in six weeks. The model works out that it needs contact search plus timeline data, and goes and gets it.

The three ways into Workspace do different jobs:

What does it feel like to use?

About 80% of the value is not switching tabs.

You’re drafting something: a listing description, an email to a client, a plan for the week. You need to know what’s actually on your calendar tomorrow, or whether that seller ever replied. Normally that means leaving what you’re doing, opening the CRM, finding the record, reading it, coming back and remembering what you were saying.

With the connector wired up, you ask, and the answer arrives in the conversation you were already having.

The other 20% is questions you’d never bother opening a CRM to answer. “Who’s gone quiet this month that I should probably call?” is a genuinely useful question that almost nobody asks their CRM, because building it as a filtered smart list is more work than the answer seems worth. Asking it in a sentence is not.

How do you connect it?

With a sign-in or a key. Both take a couple of minutes.

Claude Code. One command, using the same syntax Anthropic documents for any remote server:

claude mcp add --transport http brokeragent \
  https://mcp.brokerandagent.com/mcp \
  --header "Authorization: Bearer bak_your_workspace_key"

Or add it without the header and run /mcp inside Claude Code to sign in with OAuth instead.

Claude’s web and desktop apps. Add https://mcp.brokerandagent.com/mcp as a custom connector and sign in. Anthropic’s help center says custom connectors are available on its Free, Pro, Max, Team and Enterprise plans, with Free users limited to one.

Other tools. AI tools that support MCP over HTTP connect the same way. The protocol is open, and we didn’t invent any of it.

The API key comes from Settings → Developer, where an owner or admin can create one. Connected apps are listed in Settings → Integrations, so you can see what has access. If you can use OAuth, prefer it: there’s no key to copy, paste or leave in a file.

What can it do, and what can’t it do?

It reads generously and writes narrowly. Of its 25 tools, 20 only read.

Reading is low-risk, and it’s where most of the value is. The writes are the kind of light capture you’d want to jot down without opening anything.

The things in the last column are absent on purpose, not on a roadmap. An AI working on your behalf shouldn’t be able to do the things you’d want to look at first, and “send a message to a client” is squarely in that category however good the model is. There’s no privileged back door, either: the connector works inside the same permissions as the rest of your workspace.

How should you judge any AI connector?

By what it can change, how it signs in, and whether you can see it. The same questions apply to ours and everyone else’s.

• What can it write? Read-only access is low-stakes. Anything that sends, deletes or changes stages deserves a hard look.
• Can it message your clients? If yes, find out whether anything waits for your approval first.
• How does it sign in? OAuth means no long-lived key sitting in a config file. The MCP project’s security guidance recommends a “progressive, least-privilege” approach to permissions.
• Can you see and revoke it? You should be able to find every connected app in one place.
• Do you trust the server? Anthropic’s own guidance on custom connectors says “only connect to trusted servers” and warns that malicious servers can hide instructions for the model. That applies to any server you add, not just CRMs.

Where does your client data go when you ask?

To the AI tool you asked, as part of that conversation. That’s worth understanding before you connect anything.

When Claude answers “who hasn’t heard from me in six weeks,” it reads those contacts through the connector, and the results become part of your chat. For Claude’s web and desktop apps, Anthropic’s help center notes that the connection runs from Anthropic’s cloud, not from your own computer. So the data you ask about is handled under that AI provider’s terms, the same way it would be if you pasted it in yourself.

Two practical rules follow. Connect only AI tools you’d trust with client information, on an account where you control the data settings. And check whether your brokerage has a policy on client data in AI tools, because some do, and it’s easier to ask first.

How does it fit with the in-app assistant?

MCP is the open door. The in-app assistant does the heavy lifting.

The Workspace assistant lives beside everything in the app, with a taste every month on Free and the full pool on Pro. It can move stages, update deals, set up automations and draft messages, and the riskier actions wait for your tap before they happen. Messages to clients are drafts until you send them.

The MCP server is on every plan, Free included, and it’s deliberately smaller. Our guess is that reading and asking move almost entirely to whichever assistant you already use, while the actual doing (building the CMA, working the pipeline, running the transaction) stays in the app, because those benefit from a real interface.

Where is this going?

Toward the assistant you already use being the front door to your business software. That seems safe to say, and the rest is a guess.

Other real estate platforms are arriving too. Rechat, for example, announced an MCP server in August 2026 for working with contacts, marketing and transactions from Claude and ChatGPT. That’s good. It’s an open standard, and the whole point is that everyone can implement it.

What we’re less sure about is how much of the work moves there. For now, the useful rule is simple: let outside assistants read freely and write carefully, and keep anything that reaches a client behind a human.

If you build things and want to poke at it, the MCP server and the API work on Workspace Free, because gating the developer surface behind a paid tier is how you end up with no developers.