Skip to content

What AI assistants can do

Connect an AI assistant like Claude or ChatGPT to our MCP server and ask it to do real work on a store’s catalog: check which vehicles a product fits, add or remove fitment, pull a coverage report, export a filtered CSV, or fix the words shoppers search for but the catalog doesn’t use.

I want to…Ask something likeSection
See what vehicles a product fits”What does SKU 24-186766 fit?”Check and fix vehicle fitment
Add missing vehicles to a product”Add 2018-2021 Jeep Wrangler to SKU 24-186766”Check and fix vehicle fitment
Remove wrong vehicles from a product”Remove all F-150 fitments from SKU 24-186766”Check and fix vehicle fitment
Know how good my fitment data is”What’s my fitment coverage for Bilstein?”Measure fitment coverage
Get a spreadsheet of products”Export all products with no fitment data”Export your data
List every make and model in my catalog”Export all makes and models we cover”Export your data
See which vehicles shoppers pick”Which vehicles did shoppers pick most last month?”Read reports
See how shoppers engage with products”Which products do shoppers see often but rarely open?”Read reports
See search activity over time”Show daily searches, add-to-carts and orders for last week”Read reports
Find products to cross-sell”What do customers buy together with SKU 24-186766?”See what’s bought together
Fix search results for a term”Shoppers search ‘tpms’ but get nothing, fix it”Manage synonyms

Read the vehicles a product fits, check whether a vehicle exists in the fitment data, and add or remove fitments. Useful for a supplier’s updated application list, a customer asking “does this fit my truck?”, or a product showing No Fitment Data that shouldn’t.

Try asking
What vehicles does SKU 24-186766 fit?
Add 2018 through 2021 Jeep Wrangler Rubicon to SKU 24-186766.
Here's an application list from our supplier. Add all of these to product 7823419023.
[paste the list]

What you get back. For a lookup: the product’s fitment type (Vehicle Specific, Universal Fit, No Fitment Data, Non-Vehicle Products, or Many Fitments) and its vehicles, with consecutive years grouped into ranges like 2018 - 2020. For a write: how many fitments were added. Vehicles the product already had are skipped.

Good to know:

  • Identify the product by product ID, SKU, or URL handle (Shopify only). Use one of the three, not several. SKU lookup isn’t available on every platform, and a SKU matching more than one product returns an error. Use the product ID in both cases.
  • 2018 - 2021 is expanded into one fitment per year.
  • Vehicles are checked against the fitment data first. If any vehicle in the request isn’t found, nothing is added, and the assistant shows the valid values to pick from. A vehicle that isn’t in the data yet can still be added once you confirm it.
  • A year range is checked year by year, so every year in it must exist in the data.
  • Ask for the valid values of a field to check spelling, such as “which models are there for a 2020 Ford?”
  • Additions show up immediately when you ask about the product again. Shoppers see them after the next index rebuild.

Remove fitments from a product, one product at a time, identified the same way as above.

Try asking
Remove all F-150 fitments from SKU 24-186766.
Take 2019 and 2020 off the Jeep Wrangler fitments on product 7823419023.
Clear all fitment data from SKU 24-186766.

Good to know:

  • You describe what to remove as a pattern, not an exact row. A field you leave out matches any value, so “remove the F-150 fitments” removes every F-150 the product has, whatever the year or submodel. Ask to see the product’s fitments first if you aren’t sure how wide that reaches.
  • Years are split, not dropped whole. Taking 2019 - 2020 out of a 2018 - 2022 fitment leaves 2018 and 2021 - 2022.
  • Fitments that came from an external feed can be removed too, not just ones added in the editor.
  • Clearing a product’s fitment data entirely shows you a preview first and needs an explicit confirmation.
  • A pattern that matches nothing removes nothing and tells you so, rather than failing quietly.
  • Removals are recorded in Fitment Operation History, tagged MCP with your email.

How much of the catalog has usable fitment data, either across everything or scoped to a vendor, category, price band, or search query.

Try asking
How's our fitment coverage?
What's the fitment coverage for Bilstein products?
Compare fitment coverage for Bilstein and KYB.

What you get back. Total product count, a breakdown by fitment type with counts and percentages, and an overall coverage percentage. The numbers come back in the conversation rather than as a file, so follow-up questions work.

Good to know:

  • Vehicle Specific, Universal Fit, and Many Fitments all count as covered. Non-vehicle products count toward the total but not as covered, so they lower the percentage.
  • Requires fitment classification to be configured for the store.
  • Scope it by naming the vendor or category you want.
  • For coverage as a table broken down by vendor, ask for an export instead.

A download link for a CSV of products, of fitment values (every make, model, and year covered), or a coverage pivot table.

Try asking
Export every product with no fitment data.
Export all products from Bilstein under $200.
Export a table of fitment status broken down by vendor.

What you get back. A download link, and for product exports the number of products matched. If the filters match nothing, you get the values that do exist for the fields you filtered on, so a near-miss like “Bilstien” is easy to correct.

Good to know:

  • Treat the link as a password. It carries the store’s export credential, so don’t post it anywhere public.
  • Product CSVs have one row per product and per fitment and per variant, so the file is larger than the product count suggests.
  • Fitment exports nest fields in the order you ask for them. Makes then models gives every make with its models.
  • A pivot table needs a row field and a column field. “Fitment status by vendor” gives vendor rows and fitment-status columns.

Ask how shoppers use search on the store over any date range. Dates are inclusive and in UTC, and long results come in pages.

Try asking
Which vehicles did shoppers pick most in October?
Show the top 10 Ford models shoppers picked from July 1 to September 30.

What you get back. The vehicles shoppers picked, with the number of visitors for each. Can be sorted by visitors (default) or any vehicle field, and searched by vehicle.

Try asking
Which products got the most impressions from March 1 to May 31?
Which Bilstein products got the most add-to-carts this year?
Show the products shoppers saw often but rarely opened in October.
Which products do 2025 Ford F-150 shoppers look at most?

What you get back. The products shoppers saw or opened, each with its ID, SKU, name, brand, and URL, and for each one:

ColumnMeaningSortable
ImpressionsThe number of unique users who saw the product in search results.Yes (default)
Page Visit UsersThe number of unique users who opened the product page.Yes
Page Visit RatePage Visit Users as a percentage of Impressions.Yes
Add to Cart UsersThe number of unique users who added the product to cart.Yes
Add to Cart RateAdd to Cart Users as a percentage of unique users who saw the product in search results or opened its page.Yes
Confirm Order UsersThe number of unique users who placed an order with the product.Yes
Confirm Order RateConfirm Order Users as a percentage of unique users who saw the product in search results or opened its page.Yes
Order AmountThe amount paid for the product in confirmed orders (price × quantity).Yes

Can be searched by name, SKU, brand, or ID, and narrowed to one brand or one vehicle.

Try asking
Show daily searches, add-to-carts and orders for last week.
Show monthly active users and orders for this year.
Which hours yesterday had the most searches?

What you get back. One row for each hour, day (default), or month with search activity. Hourly stats cover at most 31 days.

ColumnMeaning
Active UsersThe number of unique users who searched, clicked a result, saw a product in results, added to cart or ordered.
SearchesThe number of search and autocomplete requests, including category pages, filter changes and paging.
Unique QueriesThe number of distinct search phrases users typed.
Search Click UsersThe number of unique users who clicked a product in search results.
Search Click RateSearch Click Users as a percentage of Active Users.
Add to Cart UsersThe number of unique users who added a product to cart.
Add to Cart RateAdd to Cart Users as a percentage of Active Users.
Confirm Order UsersThe number of unique users who placed an order.
Confirm Order RateConfirm Order Users as a percentage of Active Users.

The products customers bought in the same orders as a product, useful for choosing what to recommend alongside it. Shopify stores only.

Try asking
What do customers buy together with SKU 24-186766?
Which products bought with SKU 24-186766 fit any of the same vehicles?

What you get back. Up to 10 products, most often bought together first, each with its ID, SKU, and name.

Good to know:

  • Identify the product by product ID, SKU, or URL handle, the same way as for fitment.
  • The products come from the store’s orders in the last 90 days and are refreshed about every 3 days.
  • The products aren’t filtered by vehicle. To keep only those that fit the same vehicles as the product you asked about, ask the assistant to compare fitments.

The sets of words search treats as equivalent, so a shopper searching tpms also finds products described as tire pressure sensor. Groups can also carry exclusions, for when a term pulls in products that don’t belong.

Try asking
Do we have a synonym for "tpms"?
Shoppers searching "tpms" get nothing. Set up a synonym with "tire pressure sensor" and "tire pressure monitoring sensor".
Searching "wheel" brings back steering wheels. Can we exclude those?

What you get back. The matching groups with their terms and exclusions, or a confirmation of your edit plus any warnings.

Good to know:

  • Many stores inherit a shared synonym list. Editing requires a store-specific copy first, either copying what is inherited or starting empty. Going back to the inherited list discards those changes and asks you to confirm.
  • Write terms as plain words. Multi-word terms like brake pads need no quotes; a quoted term only matches when a shopper searches with quotes.
  • Exclusions only apply when the query matches the group’s first term, so put the term you’re fixing first. They also require the negative-synonyms feature, and you get a warning if it isn’t enabled for the store.
  • Edits can be validated without saving, and deletions show you the group before removing it.
  • Name products precisely. A SKU or product ID beats a description.
  • Paste raw data. Supplier lists don’t need reformatting first.
  • Ask follow-ups. The data stays in the conversation: “now break that down by vendor”.
  • Scoped to one store, the single store whose server URL you connected.
  • It reads catalog data, fitment data, and store analytics reports.
  • It can write vehicle fitments and synonym groups. Nothing else: not products, prices, inventory, search configuration, the Convermax account, or the ecommerce platform’s admin.
  • Fitments are edited one product at a time. Bulk changes across many products still belong in the Fitment Editor.
  • One unknown vehicle blocks the whole add. A vehicle you confirm is added whether or not it exists in the data, so skim large pasted lists.
  • Synonym deletions take a second confirmation and show you the group first.
  • Fitment changes reach shoppers at the next index rebuild, not immediately.
  • Export links carry a credential. Anyone with the link can download that data.

Review and revoke connected apps from the MCP Server card on the Integrations page in the Convermax app. See Managing connected apps.

We’re expanding what AI assistants can do with a store.

Have a workflow you’d like to hand off to an assistant? Tell us at support@convermax.com.