For developers

An MCP server. Run audits from inside Claude, Cursor, or any MCP client.

Screenshot to comeAn MCP client calling get_findings on a store’s audit, each finding with its confidence as data.
A real screenshot from our own WooCommerce test store goes here once it’s built.
Why this exists

Your developers already live in these tools.

Switching to a browser to run an audit, then back again to act on it, is enough friction that it doesn’t happen.

How it works
1

Generate a key

In Settings, shown once.

2

Connect

Your client to the endpoint, with that key.

3

Nine tools

Audit any public store, list and run audits on your own stores, read findings, read a product’s issues, generate fixes, export them.

4

Every finding carries its confidence

As structured data, not just prose.

Read this before you rely on it

What it can’t do.

It never publishes.

There’s no publish tool, and there won’t be. Publishing stays in the web interface where a person looks at each field. An agent acting on your catalogue isn’t a person reviewing it.

It can’t delete a store, change credentials, or touch billing.

None of those are exposed.

It’s scoped to your own account.

Another account’s stores return the same answer as ones that don’t exist.

Generating fixes on our AI uses your credits.

The same credits as the web, and a run on ours waits for you to confirm it in How’s My Store first. On your own AI key, it doesn’t count. Free audits of any public store are capped per key each day, like the website’s free check.

What you get
  • Nine read-and-propose tools
  • Confidence and measure on every finding, as data
  • Per-account keys, revocable
  • The same numbers the web interface shows, because it’s the same stored report

Start free.

The MCP server comes with some plans — Pricing shows which. Then Settings → MCP access to make your key.