Storewalk in Claude
You can ask Claude how ready your store is for AI shopping agents and get an answer from a real Storewalk scan. Add the Storewalk connector from Claude's connector directory, give Claude your store's address, and it runs the same scan as the form on our home page, then explains what it found in plain words. This page is the connector's documentation, so it is a little drier than our blog. Only a little.
Last updated: 10 November 2026
What does the connector do?
It gives Claude two tools for one job, checking a store:
- start_store_scan starts a Storewalk scan of the address you give. A scan takes a few minutes, so the tool answers straight away with a scan id.
- get_store_scan takes that id and reports progress until the scan is done, then returns the results.
You never call them yourself. Ask something like “how ready is shop.example.com for agentic commerce?” and Claude picks the tools, waits for the scan and tells you what it means. For the details behind an answer, ask Claude to go through the failing checks one by one.
What do I see while it runs?
On claude.ai, the desktop app and the mobile apps, a Storewalk card appears in the chat as soon as the scan starts. It shows where the scan is on an agent’s path and each check as it lands, so you have something to watch besides the kettle. When the scan ends, the card shows the score, where an agent would stop and the most urgent results, with buttons for the full list and the full report. Claude then explains what it all means. In Claude Code there is no card; Claude tells you the progress in text.
What does Claude get back?
The same Ready results the agent readiness scanner shows on the site, written for a model to retell:
- the Ready score, or the reason we are not showing one, for example when the store’s bot protection kept our crawler off the homepage;
- the step where an AI shopping agent would stop;
- every Ready check that has a verdict, one line each: the verdict, what the check means for an agent and the first step of the fix;
- how many checks could not be measured, and why;
- a link to the full report on storewalk.ai, with the evidence, the request log and the CSV.
Security results (our Safe checks) are not part of it. On the site they need the domain owner to verify first, and that does not change because the question came through a chat.
What does a scan read?
Public pages only, the same way the scan on our site does. Our crawler identifies itself, obeys the store’s robots.txt and keeps to a polite pace. It does not log in, add anything to a cart or try to get past bot protection. If a store turns the crawler away at the homepage, the result says so and covers what sits outside that wall. The bot page documents the crawler and how to let it in.
What are the limits?
They are ours, and they protect the stores being scanned:
- One new scan of a store an hour. Scans started on storewalk.ai count as well as those from Claude.
- Recent results are reused. If the same store was scanned from Claude in the last 24 hours, you get that scan. Ask Claude to scan again “fresh” after you have fixed something.
- A daily cap on new scans started from Claude across everyone.
Every limit message says when to try again.
Can Claude read a scan I started on the site?
No. A scan started on storewalk.ai stays on storewalk.ai, and its full report opens with your email, as before. Claude can read only scans started from Claude, and an id from the site gets the same “not found” as an id that does not exist.
What do you keep?
From a Claude scan we receive the store address you asked about, and we keep the scan the same way as one started on the site. We do not receive your conversation, your name or your email. The scan is about a public store, not about you, so for a day its results go to anyone who asks Claude about the same store. The privacy policy has the details, and using the connector means accepting our terms.
Who do I contact?
Write to info@storewalk.ai for support, a question about a result, or a security issue in the connector. We read everything, including the long ones.