Skip to content
Scope an integration

Glossary

What is the Model Context Protocol (MCP)?

The Model Context Protocol (MCP) is an open standard that lets AI assistants and agents connect to external tools and data through a common interface.

The Model Context Protocol (MCP) is an open standard, introduced by Anthropic in 2024, for connecting AI applications to external systems. An MCP server exposes tools, resources and prompts in a standard format, so any compatible AI client can discover and use them, for example to look up records or take actions in a SaaS product.

How does it work?

  1. The SaaS vendor builds an MCP server that wraps its API.
  2. The server describes available tools (actions) and resources (data) with names, descriptions and input schemas.
  3. An AI client connects, authenticates the user, lists the tools and calls them as needed.
  4. The server enforces permissions and returns results the model can use.

MCP is designed so that one server can serve many clients. For a SaaS vendor, that means building on top of a well-documented API with clear authentication, rather than creating a separate integration for every AI product.

Where it shows up in integrations

  • Agent-ready products Customers can ask an AI assistant to look up or update data in the product.
  • Internal copilots Teams connect their own assistants to HR, LMS or CRM data through MCP servers.
  • Integration coverage One MCP server works with several AI clients instead of a custom plugin for each.

Common pitfalls

  • Exposing broad write actions without scoped permissions or confirmation steps.
  • Vague tool descriptions, which lead the model to pick the wrong tool.
  • Passing user data to tools without audit logging.
  • Treating the MCP server as a replacement for a well-designed public API rather than a layer on top of one.