https://agenticcart.ai/mcpThe Model Context Protocol (MCP) is an open standard that lets an AI assistant call outside tools. This server gives an assistant five tools that read a live store, product page, feed or llms.txt and grade it, in the middle of the conversation. Everything it reads is public, so it works on any store on the web, including ones you do not own.
Connect it
Add the server. In Claude Code, one command:
claude mcp add --transport http agenticcart https://agenticcart.ai/mcpOr put it in a config file. For clients that list their MCP servers in a JSON file:
{
"mcpServers": {
"agenticcart": { "url": "https://agenticcart.ai/mcp" }
}
}Ask. There are no commands to learn. Say what you want and the assistant picks the tool:
Check whether examplestore.com is ready for AI shopping assistants.Five tools, one endpoint
Each returns a score from 0 to 100 with specific findings, not a pass or fail. Each takes one address.
- Audit a whole store
check_store_readiness - Scores a store from its homepage address: whether AI crawlers are allowed in robots.txt, whether product data is in the pages, sitemap, llms.txt and HTTPS.Takes
url, for examplehttps://examplestore.com - Audit one product page
check_product - Scores one product page for how well AI shopping assistants can read it: product data, price and availability a machine can read, images, description, GTIN and brand. Says so when the page only gets its content from JavaScript.Takes
url, for examplehttps://examplestore.com/products/trail-shoe - Grade a feed for ChatGPT
check_acp_feed - Grades a product feed against the Agentic Commerce Protocol and OpenAI's product feed specification, field by field across the sampled products.Takes
url, for examplehttps://examplestore.com/feed.xml - Validate a Google Merchant feed
check_merchant_feed - Validates a product feed against the Google Merchant Center product specification: required attributes, price format, availability values and GTIN check digits.Takes
url, for examplehttps://examplestore.com/google-feed.xml - Check an llms.txt
check_llms_txt - Reads a site's /llms.txt and checks it against the llmstxt.org format: title, summary, link sections, whether the listed links answer, and whether the server sends the file rather than an HTML page.Takes
url, for examplehttps://examplestore.com
What comes back
Findings grouped by what needs fixing, what is a warning and what passes. This is an example of the shape, not a real store:
# Check store AI readiness: examplestore.com
Score: 40/100 (grade fair)
8 findings: 2 needs fixing, 1 warnings, 5 passing.
## Needs fixing (2)
- **Organization / store schema**: No Organization schema on the homepage. …
- **llms.txt**
## Warnings (1)
- **AI crawler access (robots.txt)**
## Passing (5)
- **Sitemap**
- **Served over HTTPS**
- …When a page only gets its content from JavaScript, the result says so in a note above the findings. ChatGPT, Claude and Perplexity read pages without running JavaScript, so a score alone would mislead there.
Free, and complete with a plan
Every issue is named without an account, so you always see the whole picture. What to change for each one comes with any AgenticCart plan, through the licence key that is already in your dashboard.
| Without an account | With any plan, from $29.99 a month | |
|---|---|---|
| All five tools | Yes | Yes |
| Score and grade | Yes | Yes |
| Every issue named, grouped by severity | Yes | Yes |
| What to change, for one issue as a sample | Yes | Yes |
| What to change, for every issue | No | Yes |
| The full result as data for your agent | No | Yes |
| Checks per hour | 10, per IP address | 300, per account |
Already a customer? Send the licence key from Dashboard > API Key as a Bearer token and every check comes back complete. See the plans or read the MCP documentation.
On Shopify? “Shopify MCP server” means three different things; our guide Shopify MCP servers, explained sorts them out.
Prompts to try
Copy one, put in a real address, paste it into your assistant.
Audit a store
Use AgenticCart to check whether examplestore.com is ready for AI shopping assistants, then summarise the three most important fixes.
Compare stores
Run the AgenticCart store readiness check on these three stores and put the scores in a table with the biggest gap for each.
Debug a feed
Grade my product feed at examplestore.com/feed.xml against the Agentic Commerce Protocol and tell me which fields are missing.
Fix a product page
Check examplestore.com/products/trail-shoe with AgenticCart and write the JSON-LD I need to add to fix what it found.
Check while building
I just changed the product page. Run the AgenticCart product check again and tell me what changed in the score.
Check the llms.txt
Check the llms.txt of examplestore.com with AgenticCart and tell me what to fix first.
Frequently asked questions
Do I need an AgenticCart account?
Not to try it. Without any credentials you get the score and every issue named and grouped by severity, plus what to change for one of them as a sample. What to change for the other issues comes with any AgenticCart plan. All five tools work on both; the difference is the detail, never the diagnosis.
Which AI assistants can use it?
Any client that connects to a remote MCP server over HTTP. We test it with Claude Code. It is a remote endpoint, so there is nothing to install and nothing running on your machine.
What can it check?
Five things: a whole store from its homepage, a single product page, a product feed against the Agentic Commerce Protocol, a product feed against the Google Merchant Center specification, and a site's llms.txt. Each returns a score from 0 to 100 with specific findings rather than a pass or fail.
Does it work on stores I do not own?
Yes. Everything it reads is public: the pages, the robots.txt, the sitemap and the feed. An agency can audit a client's or a competitor's store before it has any access.
What are the limits?
Ten checks per hour without a key, counted per IP address. With a plan it is 300 per hour, counted per account, so a team behind one shared connection does not share one budget. Reaching the limit is not an error in your request: your assistant is told to wait and can try again. Both limits are separate from the ones on the tools on our website.
Can it change anything in my store?
No. The server only reads. It fetches public addresses and analyses them. It cannot write to a store, cannot read anything behind a login, and has no access to any AgenticCart account, catalog or conversation. Requests aimed at private or internal network addresses are refused.
Is it the same as the checkers on your website?
The same checks and the same scores. The difference is where the result lands. A web form gives you a report you then act on by hand; a tool result lands in the conversation with everything else your assistant can do, so it can read the findings, explain them and write the fix.
Why is my score low even though my store looks fine?
A store that looks fine to a person can be unreadable to an assistant. The usual causes: AI crawlers blocked in robots.txt, no product data in the HTML, pages that only get their content from JavaScript, no llms.txt, and feeds without GTIN or MPN. None of these show in a browser.
How do I get the full findings?
Take the licence key from Dashboard > API Key in your AgenticCart account and send it to the MCP server as a Bearer token. Every plan includes it; there is no separate MCP subscription. If the key is missing or wrong, or the subscription is no longer active, the server returns the free result and says why instead of failing.
Can I use it in my own agent?
Yes. It is a standard MCP server over HTTP: JSON-RPC 2.0 by POST, with the methods initialize, ping, tools/list and tools/call. Any client or framework that speaks MCP can connect to the same endpoint.
Checked it. Now fix it across your catalog.
Enter your shop’s address. The scan shows what ChatGPT, Claude and Gemini can read today and what AgenticCart prepares for them.
Or start the free 7-day trial right away.