Glossary

What is an MCP connector?

An MCP connector is a small server that exposes one tool's data and actions to Claude over the Model Context Protocol (MCP), the open standard for connecting AI to outside systems. It gives Claude a defined set of tools to read records and, where allowed, create or update them inside your existing account.

Build my connector Browse connectors

Why does an MCP connector matter?

An MCP connector matters because it turns Claude from a chat window into a system that can use your software. Without a connector, Claude only knows what you paste in. With one, Claude can pull live data from your tool and take real actions in it, so the work happens where your records actually live.

How does an MCP connector relate to a connector in Claude?

A connector in Claude is the MCP connector you add and turn on. The Model Context Protocol is the shared standard, and an MCP connector is one server built to that standard for a specific tool. When you add a connector in Claude, you are pointing Claude at that server and the tools it exposes.

Off-the-shelf connectors exist for many popular apps. Custom, in-house, or heavily modified systems are not in Claude's directory, so a connector has to be built for them. You can browse the tools we build MCP connectors for to see which systems are covered.

What does an MCP connector look like in practice?

In practice, an MCP connector is a server that maps your tool's API into named tools Claude can call. A CRM connector, for example, might expose a tool to search contacts, one to read a deal, and one to update a field. Claude picks the right tool for your request and runs it against your account.

Read access lets Claude pull and summarize what is already in your system. Write access lets it create and update records, while destructive or outbound actions stay gated behind a confirmation you control. For the full picture of how connectors are built, scoped, and maintained, see the complete guide to custom Claude connectors.

Related terms

MCP (Model Context Protocol) the open standard an MCP connector is built on.
Custom connector an MCP connector built for a tool that is not in Claude's directory.
Full read-write a connector that can both read records and write them back, with destructive actions gated.

Frequently asked questions

Is an MCP connector the same as a Claude connector?

In practice, yes. When people say Claude connector they usually mean an MCP connector: a server built on the Model Context Protocol that lets Claude read and act on an outside system. MCP is the open standard underneath, and a connector is one server built to that standard for a specific tool.

Does an MCP connector let Claude write data, not just read it?

It can. An MCP connector exposes tools, and a tool can read or write. A connector can expose write tools so Claude creates and updates records, while destructive or outbound actions like deleting or sending stay behind a confirmation you control. A connector never grants Claude more access than the connected account already has.

Do I need to build my own MCP connector?

Only if a connector for your tool does not already exist or does not do what you need. Popular apps have off-the-shelf connectors. Custom, in-house, or heavily modified systems are not in Claude's directory, so a connector has to be built for them. That custom build is the work we do for you.

Where does an MCP connector run?

An MCP connector runs as a small server that sits between Claude and your tool's API. We build and host that server, so you do not run or maintain any infrastructure. It authenticates as your account, using OAuth where the tool supports it or a scoped key where it does not.

Need an MCP connector for your tool?

Book a short scoping call. We'll map your system's data and actions and build the connector that gives Claude scoped read-write access.

Build my connector Talk to us