Build on StoreConnect using AI agents

A StoreConnect store on v22 or later can be built, designed, and configured by AI, using the StoreConnect Model Context Protocol (MCP). Use this article to understand what that involves before you set it up.

The MCP, together with StoreConnect’s downloadable AI skills toolkit, acts as a translation layer between an agent and your store. Once it is set up, you describe what you want in plain language, and the agent proposes the changes for you to approve before they go live on your storefront.

There is no tool lock-in. You can connect a preferred agent, such as Claude, Codex, Gemini, GitHub Copilot, or Cursor. You can also install the skills on their own, without connecting anything: see Install the StoreConnect AI skills.

How it works

Three things work together when an agent builds in your store:

• The MCP server — the authorized gateway between the agent and your store. It is scoped to a single store, and an agent works only within the permissions of the StoreConnect user it signs in as.
• The AI skills toolkit — a curated GitHub repository that teaches an agent how to work with StoreConnect, including how to find information and how to make changes.
• Documentation — StoreConnect’s docs provide the step-by-step recipes an agent follows to perform an action.

Because the MCP server belongs to your store, the agent you connect only ever works in that store. MCP requires StoreConnect v22 or later, so check your store’s version before you set anything up. To get connected, see Connect an AI agent to your store.

What agents can do

An agent can do anything in StoreConnect that its StoreConnect user has access to do. The best-covered areas are the ones specific to StoreConnect rather than Salesforce: store design and themes, and working with Liquid and custom objects.

You prompt for changes in plain language, without needing to know the technical detail behind StoreConnect. Tested scenarios include:

• Build a complete storefront from an existing website, for example recreating a Shopify design in StoreConnect.
• Create and update page content, sections, and media.
• Set up StoreConnect for the first time: install it, add a sync user, and complete the basic configuration.
• Bulk migrate media and other data, pre-mapped to StoreConnect custom objects.
• Launch a sale: update page banners and promote discount codes.

Working safely with an agent

An AI agent acts with real effect in your store, so treat it as you would any tool with write access to your data. The following guardrails are in place:

• Permissions — an agent signs in as a StoreConnect user and can only do what that user is allowed to do. Grant it only the access it needs for the task.
• Approval — by default every change is staged for you to review, and nothing reaches your store until you approve and then publish it, much like changes work now in the website builder. See review and approve a content change. By default the person who publishes cannot be the user the agent signed in as, so a solo operator needs the Publish Own Changes scope on their role. Grant it only where you intend it.
• Store scope — an agent only ever works in the one store you connect it to. Connecting a different store means using that store’s own MCP server.

:::warning Try agent-driven changes in a test store or Salesforce sandbox before you run them against your live store. See Set up your test store and Move data from sandbox to production. :::

Next steps

When you are ready to connect an agent, work through Connect an AI agent to your store, then Install the StoreConnect AI skills. If you are the administrator deciding who may connect one, start with Manage who can use an AI agent.

For jobs with enough moving parts that the order of operations matters, see Agent recipes.