Back to all posts
Security Published October 14, 2026 4 min read

Scan a Shopify Store for Risky Scripts From Claude, ChatGPT, or Cursor

What it looks like to ask an AI agent to scan a Shopify store over MCP, what comes back, and what a person still has to decide.

By The Appnary Team

MCP, the Model Context Protocol, is a way for an AI agent to call tools in another system. A store scanner is a good fit: the agent asks for a scan, and the scanner returns findings in a form the agent can read.

What the agent can ask for

Vigil's MCP tools are scan_store, warm_live_page, block_host, unblock_host, and list_blocked_hosts. A per-store agent token controls which store a connection can reach. In testing, you create the token for a store yourself.

  • scan_store returns findings for the published theme, app embeds, app blocks, and external frames.
  • warm_live_page loads the rendered storefront so scripts added at runtime can be seen.
  • block_host, unblock_host, and list_blocked_hosts turn a script host off on the storefront, undo that, and show what is currently off.

What comes back

Each finding names the script, its source (a theme file and line, or an app), and the reason it was flagged. A flagged secret is reported without printing the secret. Because the output is structured, an agent can sort it, group it by app, and explain it in plain words.

What stays with a person

An agent can read the findings and draft a plan. The decision to turn a script off is a merchant's. A script can be one you need, and blocking it can break a checkout flow or a tracking setup. Vigil blocks scripts that load after its blocker embed, and for scripts hard-coded in the theme head it flags the line to remove instead. The first version does not uninstall apps, and it does not read customer records or order contents.

Where this stands

This is in testing on development stores. Public MCP access for every store, and the Shopify App Store listing, are planned and not available. If you want a manual version of the same check today, start with how to find which app added a script.

Read the Vigil notes or join the waitlist. There is no install link until Shopify approves the app.

MCPShopify securityAI agentsVigil

Frequently Asked Questions

What is MCP?
The Model Context Protocol is a standard way for an AI agent such as Claude, ChatGPT, or Cursor to call tools in another system and read structured results.
Can an agent turn off a script by itself?
Vigil has a block tool, but the decision is the merchant's. The first version does not uninstall apps.
Can I use this today?
Only in testing on development stores with a per-store token. Public access and the App Store listing are planned.