MCP Connector

Simple Definition

An MCP connector is the practical connection that lets an AI app or agent use a specific external tool or data source through the Model Context Protocol (MCP).

Instead of every AI tool needing a completely custom integration, MCP connectors give agents a more standard, structured way to reach files, databases, code, docs, and business apps.

What an MCP Connector Can Do

• Read files or search a codebase
• Query a database
• Inspect project docs or issues
• Pull context from a connected app
• Trigger an approved action through a defined interface

A connector usually runs as an MCP server that exposes those tools, while the AI app acts as the MCP client that calls them.

Example

An MCP connector could let Claude or a coding agent read your project files, inspect documentation, or query a database through an approved interface, without you copying and pasting everything into the chat.

Why Permissions Matter

A connector that can only read data is very different from one that can send, delete, or change things. Good setups start read-only, use least privilege, and keep a human in the loop for risky actions.

Related Terms

• Model Context Protocol, the open standard MCP connectors are built on
• AI Connector, the broader idea of connecting AI to tools
• AI Agent Infrastructure, where connectors fit in the bigger picture
• Tool Use, how AI calls external tools
• AI Permission Hygiene, managing what AI can access