Shopify Editions Spring '26: Sidekick (All 9 Items)

Until now, Sidekick could basically only operate and answer questions about Shopify's own native features (products, orders, analytics, and so on). Even if you asked "How are my Klaviyo emails performing?", it couldn't answer, because Klaviyo is a third-party app. This time, a mechanism for third-party apps to "plug into" Sidekick (Sidekick app extensions) has been introduced, making Sidekick feel even more like a "hub."

• Data integration (answering questions): Sidekick searches an app's data to respond. Example: "Which email subject lines performed best?" → it searches Klaviyo's data and answers.
• Action integration (performing operations): it guides you to the correct screen inside the app, or displays a dedicated UI right there to operate. Example: "Edit this email" → it takes you straight to the edit screen for the relevant campaign.

Marketing: Klaviyo / TikTok / Seguno / Mailchimp ／ Customer management: Loop Returns / AfterShip / Judge.me / Smile / Yotpo ／ Promotions: Abra / Discount Kit ／ Products & operations: Printful / Syncee / Matrixify / Avia.
→ The latest list of supported apps can be checked in the official story (Apps that integrate with Sidekick).

App developers add a Sidekick extension to their own app and define "tools" that Sidekick can call. Tools come in both kinds: retrieving data (e.g. answering "Which email subject lines performed best?" with Klaviyo data) and performing operations (e.g. replying to a review / creating a product review), and for each one you write a description of "what the tool does and what input it takes."

The crux of the mechanism is routing. Users don't need to specify an app name. Sidekick reads the "extensions summary" each app registers along with the description of each function, and automatically decides "which app should handle this question or operation" before handing it off. Think of Sidekick as the "front desk (orchestrator)" that bridges to the right app behind the scenes.

At runtime, tools run in Shopify's sandbox (an isolated, secure execution environment) with the same permissions as that app. Technically it's a method similar to MCP (Model Context Protocol, a standard for providing tools to AI), but it's Shopify's own implementation, and existing MCP servers cannot be reused as-is, according to the docs.

1. Add a Sidekick extension to the app.
2. Define tools: for each of data retrieval and operations, write a description explaining its purpose and input.
3. Write an extensions summary (within 256 tokens) — this is the foundation of Sidekick's routing decisions. Each function's description is also used for routing.
4. Implement the tools so they run in the Shopify sandbox under the app's authentication and permissions.

Constraints: up to 5 intents and 20 tools per app. As a rough guide for response time, data retrieval ≈ 1 second / operations ≈ 3 seconds. See the developer docs for details.

The fragmentation of "a separate dashboard for each of 20 apps" is being consolidated into a single conversation. The flip side is that apps that don't build an extension may become harder to see in a merchant's day-to-day workflow. Sidekick is becoming the "operation hub for your apps" — this is the most strategic item this time, sitting right at that entry point.

For convenience in this article, we'll frame Sidekick itself as something you "ask questions of (pull-type)," and this feature as something where "Shopify proactively makes suggestions to you (push-type)" (pull/push are not terms from the source — they're labels we use to make this easier to understand). In practice it's the "home cards" lined up on the admin home, whose content changes according to your store's activity. Note that on the home screen, questions to Sidekick itself and these home cards coexist.

• Update information about new features
• Advice on making the most of Shopify
• Guides for recommended tasks (tips for boosting sales, improving CVR, and so on)
• Daily insights = your own store data is analyzed every day and observations are surfaced automatically (the highlight this time). They fall broadly into three types (comparison with the prior period / ongoing trends / shifts in best sellers).

You can hide home cards you don't need (though cards that include order tasks can't be removed). When you hide one, sending feedback helps improve the accuracy of suggestions. Being able to tune what you receive — rather than just "receiving" — is worth keeping in mind.

"Daily insights" are generated every day only for stores that have averaged 10 or more orders per week over the past 6 months. Up to 3 per day, expiring in 24 hours. They aren't generated for stores below an average of 10 orders per week — in other words, new and small stores are less likely to benefit, which is worth noting.

When we checked the home of a store with enough orders on a real device, a card with a graph reading "Sales from a certain channel rose sharply month over month. Let's identify the cause" really did appear near the bottom of the home. It came with a line chart and a "View report" button — confirming on-screen the spec that it only appears for stores meeting the scale condition.

The big shift is that AI's mode of engagement has moved from "pull" to "push." That said, the suggestions are generic, and prioritizing the right moves in your own context still requires the judgment of the merchant or their partner. Our landing point is: don't take AI suggestions at face value. And it's worth noting that this isn't "available to everyone" (the 10-orders-per-week wall).

From the Apple Watch app that syncs with the Shopify app on your iPhone, you can ask Sidekick by voice and instantly check your store's status. Answers are read aloud, and a text summary is shown on the watch screen. Example questions: "How many orders are unfulfilled?", "What are today's total sales?", "Which products are low in stock?", "How many returns are there?" and so on. Note that the Apple Watch app itself lets you constantly check sales, orders, and sessions "without asking" via watch-face complications and the home view, and Sidekick adds a "ask by voice" layer on top of that.

Its value comes down entirely to the experience of "checking in a few seconds without stopping what you're doing." It suits a quick glance on the floor — while packing or serving customers, when your hands are full — but it requires the prerequisite of having both an iPhone and an Apple Watch, so the audience it appeals to is limited. We see its importance as the "nice to have" layer.

(This behavior isn't directly documented in the source help; it's based on the Shopify Editions Spring 2026 announcement and hands-on checks in a test store.)When you give an ambiguous instruction that lacks information, instead of guessing and running with it, Sidekick asks back about the missing conditions with options (buttons) plus free text. Rather than "an AI reading the room and going off the rails," it's designed to "check with a human at the branch points."

In the test store, when we sent only "create a discount," Sidekick didn't execute and instead asked back "What kind of discount would you like to create?", presenting the options percentage off (% off) / fixed amount discount / free shipping / Buy X Get Y (purchase perks such as buy X get Y free) / other (free text). You can pick by clicking, and write non-standard requests under "other." It also asked in multiple stages — for example, "Is this for the entire order or a specific product?"

It's a modest but effective safety valve that reduces the biggest worry about delegating work to AI: "it builds something different from what I intended." It's most effective for "tasks with many settings" such as discounts, products, and emails. On the other hand, articulating "what you want to do" still rests with the human.

When you ask for a time-consuming task (data analysis, processing large numbers of products, report generation, and so on), this feature lets the processing keep running in the background even if you close the chat or window and move on to other work. The announcement page clearly states that "Sidekick keeps working in the background even when you start a new task, run multiple chats at once, or close the window." The source help also says that "(for time-consuming tasks) it notifies you when it's ready for review" (note: in our test, the deliverable appeared within the chat, and a separate bell/toast notification wasn't confirmed because the processing was short).

In the test store, we ran a heavy job — "create improvement ideas for all products" — and when we closed the chat mid-processing and reopened it about 30 seconds later, the processing hadn't stopped and had advanced to "step 1 complete → categorizing → generating improvement ideas." Furthermore, when we fired off order analysis, customer analysis, and collection analysis in separate conversations one after another, the other conversations ran independently to completion behind the one being processed — confirming parallel processing across multiple conversations.

Until now, conversational AI assumed it "only works while it's open." Keeping running even when closed = you can delegate work asynchronously, which feels like a paradigm step forward. In EC operations, where there's a lot of "long-wait processing" like analyzing all products or bulk reviews, this directly affects productivity. The benefit is biggest for mid-to-large stores with many orders and products.

If you ask Sidekick to "build an app that does X," it can automatically generate a small admin tool that runs inside the admin. For example, "an app that recommends which products to reorder based on sell-through rate," "an app that generates discount checkout links and QR codes," or "a task management tool for the team" — tools for internal operations. The point this time is the improvement of its editing features (code editing, preview, version history). Rather than generating it and leaving it as is, you can now edit and save the code directly, fix it element by element in the preview, and roll back to a previous version.

In a test store on a Grow plan or higher, when we asked it to "build a task management app," code was generated and we could review and adjust it while switching between the "Code / Preview" tabs in the app editor. In particular, the degree of freedom in editing has increased significantly. In the preview screen, you can specify elements such as buttons, titles, and display frames and give fine-grained, concrete edit instructions like "fix this part this way." Furthermore, in the code view you can actually edit and save the generated code — a big step forward considering you used to only be able to look at the preview. The version updates each time you edit, and you could check a list of past versions from a dropdown (you can also revert to an earlier version). It also supports switching between desktop/mobile display. Note that operations that change your store's data structure trigger an approval modal, and the store isn't changed unless you approve. (However, this approval check only appears during development and doesn't appear for already-installed apps. Because a generated app may change customer-facing data, the source warns that testing before installation is the user's responsibility.)

Being able to build small fixes for on-the-ground inconveniences yourself without an engineer is significant, but the eligible plans are limited. It's accurate to position this as "a powerful option for higher plans" rather than "available to everyone."

In the Shopify mobile app, no matter which screen in your store you move to, you can call up Sidekick, and it answers taking into account the screen you currently have open. You can ask by text or by voice, and on orders, analytics, products, and other pages, it stays displayed on screen no matter which page you move to (persistent) (this persistence is the behavior of the phone app; on the PC version it either moves with you or is called up via an icon). However, you can't call it up while a dialog (popup) is open on the screen. It supports iPhone and Android; tablets are not supported.

We actually tried it in the Shopify mobile app. On iPhone, a Sidekick input field stays persistent at the bottom of various screens — blog posts, product management, online store, and so on — so you can talk to it by text or voice from any screen. You can also call up the chat history right there, and on the theme editing screen it could recognize the selected element (e.g. heading text) as context and take instructions. However, the behavior differs by OS. On Android, Sidekick opens full-screen (rather than as a small persistent input field), and moreover it wasn't displayed on the theme editor screen. The fact that the look and supported screens differ between iPhone and Android within the same mobile app is something worth keeping in mind when you explain it. And, naturally, on the PC admin too it answers using the page you currently have open as context.

Even when you're not sitting at a PC, you can ask from your phone at the storefront, in the warehouse, or on the go — the crux is being able to use AI while moving around or on-site. Automatically using page context also lets you ask on the spot when you're unsure how to do something, lowering the learning cost for less experienced staff.

Sidekick auto-fills the new-customer creation form simply by telling it in natural language. The filled-in fields are highlighted in purple so you can identify what Sidekick added (per the source).

In the test store's new-customer form, when we conveyed the details of a fictitious dummy customer in a single sentence, the name, email, phone, and address were all auto-filled at once, and each field was highlighted in purple. What's more, the phone number was formatted into international format and the address was broken down into prefecture and city/ward — not a simple copy-paste, but structured. After filling, it stopped in an "unsaved" state and wasn't registered to the store until we saved.

This "auto-fill a form with words" isn't only for creating customers. It works the same way for various creation forms, such as creating and editing products, discounts, and collections, and registering a new company in a B2B store (per the source). It's best to view "customer creation" as the representative example. Note that what this feature mainly does is "creation," and for products, discounts, and collections editing is also possible. Deletion is out of scope (not mentioned in the source).

It's effective in settings with lots of manual entry — phone calls, business-card exchanges, bulk registration after events. Preventing format inconsistencies is also practical. The purple highlight makes "what the AI entered" clear at a glance, and the design that always lets a human check before saving gives a sense of reassurance.

This is a feature for verifying whether the logic is correct before activating a Shopify Flow (a mechanism that automates trigger → condition → action for orders, inventory, customers, and so on) workflow. Previously, it would reference existing data, but you couldn't test if that pattern didn't exist; now Sidekick auto-generates test events using your store's existing data.

For a workflow that "issues a discount code if the order amount is at or above a certain threshold," pressing "Generate events" auto-generated two test events — one that meets the condition and one that doesn't — based on data that actually exists in the store, such as products and collections (presented as two order events, high-value and low-value). When you select a generated event, the path of the logic it passes through (True / False) is highlighted on the flow diagram, so you can visually confirm which way it branches. Because the test is a simulation, it doesn't affect the actual store. Note that Sidekick's role is "auto-generating test events," and the verification screen itself — where you select a generated event and the path is highlighted — is an existing feature of Flow's test mode (the source example uses the "order created" trigger; behavior with other triggers needs checking).

You can prepare test events for verification in the following three ways. The workflow testing feature itself appeared in December 2025 (changelog ↗), and from the start you could select an existing event to verify the logic. The two-pattern auto-generation by Sidekick in ③ was added in May 2026 (changelog ↗), and that is the new feature this time.

1. ① Record events: monitor and capture live events that actually occur in the store in real time. Example: with the "order created" trigger, actually creating an order records that event. For when you want to use real, real-time data.
2. ② Create event: manually create a simulation event using existing store data. Example: select an existing order and use it as test data for "order created" (it isn't actually recreated). You can select and enter each field yourself.
3. ③ Generate events = this time's Sidekick: based on existing data, Sidekick auto-generates two test events, one that meets the condition and one that doesn't. No manual setup needed, and you can edit or delete them after generation.

With this ③ auto-generation, the effort of "preparing both the meets-condition and doesn't-meet-condition patterns yourself" is gone.

The biggest worry with automation is "it behaves unintentionally in production." Being able to try both cases with one click before activating quietly helps you build Flows with confidence. It pays off in operations that use Flow heavily, or when designing and delivering complex flows.