Skip to main content
MCP (Model Context Protocol) servers expose tools and data from external systems to your AI Agent. Instead of building a custom HTTP function for every action, you connect an MCP server and attach it to the agent.

​
When to use MCP

  • Browse products or customers in Shopify / WooCommerce
  • Work with payments data in Stripe
  • Reach hundreds of apps through Zapier or Make MCP bridges
  • Any vendor that publishes an MCP endpoint
For a single custom API you own, Functions & actions may still be simpler.

​
Add an MCP server

1

Open MCP Servers

Go to AI Center → AI Tools → MCP Servers.
2

Add New

Click Add New and enter the MCP server URL and authentication type required by the provider.
3

Save

Confirm the connection. Fix auth errors before attaching the server to an agent.

​
Attach MCP to an Agent

1

Edit the Agent

Open the AI Agent that should use the server.
2

Select the MCP server

In the AI tools section, select the MCP server you added.
3

Prompt for tool use

In the system prompt, describe when the agent should use store/CRM/payment tools. Clear intent language improves tool selection.
4

Test

In the Playground, ask a question that requires live MCP data (for example, “Is SKU 123 in stock?”).
Examples teams commonly connect: Zapier, Make, Shopify, Stripe, WooCommerce, and other platforms that publish MCP support.
MCP tools can read or change external data. Scope credentials tightly and treat model-selected tool calls like untrusted input on the server side.

​
Next

Functions & actions

Define custom HTTP tools when MCP is not available.

Best practices

Safe tool use and prompt patterns.