The AI assistant in Luseri

What it reads

With each question, the Assistant page sends a summary of your app's numbers. The summary is built in your browser from what the dashboard pages already show, in the language of your dashboard, so the assistant works from the same figures you see. Here is everything it contains:

A part you have not connected, such as Google Analytics or your App Store listing, is simply absent, and the summary tells the assistant so. The summary is capped at 40,000 characters.

What it answers

It explains what your numbers say and gives the two or three actions that matter most, in order, naming the figures, stores, reasons and competitors it relies on. Its instructions forbid it to invent a figure, a store, a review or a trend: when the data needed is missing, it says so and names the page or source that would provide it.

Answers are short, about 150 to 250 words, unless you ask for more. They arrive as they are written. The assistant follows the conversation: your last three questions and its answers are sent again with each new question.

It answers in the language of your dashboard, English or French. Its advice follows Shopify's rules on reviews: it never suggests incentivized reviews or asking only happy merchants. Reviews and uninstall reasons are written by other people, so it treats them as data to analyze, never as instructions to follow.

The page suggests questions to start with, such as a summary of your week or how your trials are doing. Other pages open it with a question already written: on a tracked keyword, for example, how to gain places on it against the apps ranked above you. A model can be wrong, so check a number that matters on its page.

Reply drafts for your reviews

Separately from the conversation, the Reviews page has a "Draft a reply" button on each review. The same model, Claude Haiku 5.5, writes a reply in the review's language, from the review's text, the store's name and your app's name. The draft never offers anything for changing or removing a review, and never asks for it to be changed. You check it, copy it and post it yourself in your Shopify Partner Dashboard: nothing is ever posted in your place.

Drafts have their own cap: 30 per account in any 24 hours, on top of the questions to the assistant.

Limits

• 40 questions per account in any 24 hours, or 200 for a founding member (an account created before paid plans), a cap that only stops abuse. Two shared caps also apply over 24 hours: 80 questions per IP address and 500 for the whole service. When one is reached, the assistant says so and you can ask again the next day.
• It answers signed-in accounts whose email address is confirmed and whose app numbers have already been copied from Shopify.
• A question can be up to 2,000 characters long. An answer stops at about 2,000 words.
• In the demo, the assistant gives examples written from fictional data: the AI model is not called. When the AI model is not available on the service, the page answers with calculations on your account's data instead, and says so under the question field.

What it does not do

• It does not act. It has no access to your Shopify account, your listing or your app, so it cannot change a price, post a reply to a review or run an ad. On Shopify, Luseri itself only reads.
• It only sees the summary. It does not see every store, individual payments, reviews older than the five latest, the searches merchants typed in the App Store, your in-app events beyond the health score, competitors' reviews, your code or your Shopify admin.
• It does not browse the web or open App Store pages: it works from the latest daily readings.
• It keeps no memory between conversations. A new conversation, or leaving the page, starts from zero, and Luseri does not save your questions or its answers.

What is sent to Anthropic

Each question goes to Anthropic PBC, in the United States, with the summary above and the earlier exchanges of the conversation. The summary contains names of stores (the top payers and those losing interest), uninstall reasons typed by merchants and public reviews. A reply draft sends the review's text, the store's name and your app's name.

Luseri's privacy policy says Anthropic uses this data only to provide the service and does not use it to train its models. Anthropic's privacy center says that, by default, inputs and outputs from its commercial products are not used for training. On the Luseri side, nothing of the conversation is stored: to apply the limits, the time of each question is recorded with your account and your IP address, and erased after 24 hours.

There is no API key to provide and nothing to set up: the assistant answers as soon as your app's numbers have been copied from Shopify.

Your own AI assistant, connected to the same numbers

If you already work with Claude, Cursor or VS Code, you can also connect them to your numbers, read-only, through Luseri's MCP server: create a personal token in the settings, paste the configuration into your tool, and it reads the same summary and your day-by-day series. The questions then go through the company that runs that tool, under its own terms. How to connect it.

How it compares

Several analytics tools for Shopify apps now have AI features, from listing audits to questions answered with your own API key. They are compared, as described on each tool's site, in Shopify app analytics tools compared.