Using your Model Context Protocol (MCP) servers to extend your AI agent's capabilities
AI agents use large language models (LLMs) to analyze queries that you type in a chat window and to formulate responses. You can extend your AI agent's capabilities by enabling access to the tools that are exposed in your connector-based or REST API-based Model Context Protocol (MCP) servers. These MCP tools enable the AI agent (powered by its LLMs) to connect to external applications or REST APIs and perform actions or run API operations on your behalf.
The steps to configure an AI agent to connect to an MCP server vary, so you must use a supported method to configure the connection. You will need to provide the server endpoint details to the AI agent to enable it to connect to the MCP server. Consult the documentation for your AI agent for specific instructions on how to configure MCP connections.
You can obtain the endpoint details for an MCP server by clicking its tile in the
Model Context Protocol (MCP) servers
page in the App Connect Dashboard. Then, use the Copy to clipboard icon
to copy the server URL, username, password, and HTTPS basic authentication header (if available)
from the Endpoint tab. Paste these values into the relevant configuration
format or schema that the agent supports.
When you have a successful connection to the MCP server, you can begin to use the chat window to type questions. The AI agent then executes the relevant connector actions in the connected applications or invokes the relevant REST API operations on your behalf, and displays the results in the chat window.