Model Context Protocol (MCP)

The Model Context Protocol (MCP) is an open standard for connecting AI assistants to the systems where work already happens. Instead of every product building its own assistant, a product exposes an MCP server and any MCP-capable client can use it: Claude, ChatGPT, Cursor, Codex, VS Code and others.

An MCP connection carries three kinds of capability:

• Tools: actions the assistant can call, each with a name, a description and a typed input
• Resources: data the assistant can read
• Prompts: reusable starting points the server suggests to the client

For a remote server, the client discovers the server's authentication requirements automatically and signs the user in with OAuth 2.1, so nothing long-lived is copied into a config file. Permissions are expressed as scopes the user grants at a consent screen, which is what makes read-only connections possible.

For generative engine optimization, MCP matters because the work is already split across tools. An assistant that can read your citation data, create the prompts you want tracked and log the page you shipped removes the dashboard round trip from the loop.

Prominara runs a remote MCP server for exactly this. See [Bring your agent](/mcp) for the server address and setup, or the [Connect your AI agent documentation](/docs/agents).