Customer data glossary

Model Context Protocol (MCP)

By Art Kukharenko · Published · Updated

Browse learn glossary resources

The Model Context Protocol (MCP) is an open-source standard for connecting AI applications to external systems, so they can reach data sources, tools and workflows in a uniform way.

Part of the Learn hub and the customer data glossary.

What it means in practice

The MCP documentation describes the protocol as an open-source standard for connecting AI applications to external systems and compares it to a USB-C port: one standard way to connect AI applications to data sources (such as local files and databases), tools (such as search engines and calculators) and workflows (such as specialised prompts). An AI application acts as the client; an MCP server exposes what a system can do.

For customer data, an MCP server in front of a customer data platform lets an assistant ask questions or take actions through a governed interface rather than direct database access. Tools might include previewing a segment's size or listing audiences. The risk concentrates there: the agent acts with whatever the server allows, so scope, authentication, rate limits, approvals for actions that send or change data, and logging must all be designed in. Content returned to the model can also carry hidden instructions (prompt injection), so treat tool output as untrusted input.

Where the server runs matters for regulated data. An MCP server can run inside your own network beside locally hosted models, so no customer data reaches an external AI service.

Example (retail bank): an analyst assistant can call a "segment size" tool that returns counts only above a minimum threshold, while an "export" tool is simply not exposed. Every call is logged with the user and the tool arguments.

Where it matters

Frequently asked questions

Is MCP an API replacement?

No. MCP sits above APIs. A server wraps existing APIs as tools that an AI application can discover and call in a standard way.

Is MCP safe for customer data?

It is a protocol, not a security control. Safety comes from what the server exposes, authentication, approvals and logging. Start read-only with minimal fields.

Does MCP require a cloud service?

No. Servers can run locally or inside your network, which keeps data within your boundary when paired with locally hosted models.

Back to the customer data glossary or the Learn hub.

Further reading

Published: · Updated:

Publication date recorded for the approved release. Editorial updates do not establish legal approval or a customer result.

Need help?

Ask the PrivyCDP Assistant