[Participant Report Vol. 1] I Attended Shopify DotDev 2026 - What is DotDev? Highlights from the Event

From July 21st to 22nd, 2026, Shopify DotDev 2026 was held in Toronto, Canada. This is Shopify's annual technical conference for developers and partners. Four members from our company, Flagship, attended in person. We divided up the areas of responsibility and recorded 34 sessions and conversations over two days, covering everything from keynotes to workshops and consultations at booths. Including the community event on the following day, the total comes to 43.

This report will be delivered as a series. This first installment introduces DotDev as an event and highlights some of the particularly impressive aspects we experienced there. The detailed content of each theme will be explored one by one in subsequent reports.

What is DotDev?

DotDev is a technical conference hosted by Shopify for developers and partners. It's where Shopify's new features and future direction are first announced, bringing together app developers and agencies from around the world. At this year's opening, it was mentioned that 68 countries were represented in the venue alone. Here, "agency" refers to a production company or support company that specializes in building and operating e-commerce sites using Shopify. It's helpful to think of them as entities that comprehensively support design, development, and marketing on behalf of businesses looking to launch online stores.

The program isn't just about listening to presentations. In addition to keynotes and breakout sessions, there are hands-on workshops and booths where attendees can speak directly with product team members. This allows participants to immediately test announced features or ask developers about challenges they face in their projects.

The two days had clear differences in their roles.

Date	Positioning	Main Content
Day 1 (July 21)	Product Day	New feature announcements and technical sessions from morning to evening
Day 2 (July 22)	Partner Business Day	Discussion between CEO Tobi Lütke and President Harley Finkelstein, case studies of migrations from other platforms, agency management in the AI era

This year, the focus of discussion was clearly on AI. Compared to previous attendances, the central topic has shifted significantly to Agentic Commerce, where AI agents shop on behalf of people, and it was clear throughout the two days that Shopify is steering in this direction. For our company, this business trip was an opportunity to confirm this next step firsthand and bring that information back to our projects in Japan.

What was presented to partners

The first day began with Atlee Clark's discussion on the partner ecosystem. He stated, "This room, meaning Shopify and its partner ecosystem, will define the next era of commerce," and continued that the next 10 years would be special. He then presented several concrete figures.

First, there was the app review waiting time, which had been a long-standing source of dissatisfaction. At its peak, it took up to 40 days, but it is now 4 days. This was introduced with the comment, "I've been at Shopify for 11 years, and it's never been 4 days," which drew applause from the audience. He also said that partner support inquiries can now be answered four times faster than a year ago, thanks to restructuring the channels.

The compensation model is also changing. A new model will allow partners who build and launch a store to share in the growth of that merchant's GMV (Gross Merchandise Volume). This was described on stage as "sitting on the same side of the table." The revenue share paid out last year was $1.3 billion, and it was shared that for every dollar Shopify earns, partners generate seven times that amount.

It was also mentioned that 70 partner meetups have been held worldwide in the past six months. In addition to classic cities like New York and London, Vilnius, Istanbul, Taipei, and Fukuoka were named.

Local Highlights

Here are seven particularly impressive things we saw and heard at the event.

1. UCP: The Common Rule for the Age of AI Shopping

This year's star was Agentic Commerce. The foundation for this, developed jointly by Shopify and Google, is the open Universal Commerce Protocol (UCP), which allows AI agents to transact with any store using the same methods. It was released in January 2026, received a major update in April, and the next release is scheduled for August. This wasn't a new announcement at DotDev, but rather an update on the current status of an already active initiative.

The session featured Mani Fazeli (VP Product), who co-created UCP, and Ilya Grigorik (Distinguished Engineer), who has a background in establishing standards like HTTP and TLS. They discussed the philosophy behind UCP: it was designed as an infrastructure to expand the entire ecosystem, not to enclose it. The goal, they explained, is "diffusion, not aggregation" – a myriad of commerce experiences connecting to merchants, rather than demand concentrating with a few giant players. Another striking design perspective was rethinking checkout not as a fixed procedure, but as a multi-stage negotiation where AI and the store iron out conditions. UCP is enabled by default for all merchant storefronts and requires no special configuration.

2. Catalog: A New Foundation for Partners

Another topic repeatedly discussed alongside UCP was Catalog. This is a search engine built specifically for product search, designed as something distinct from general information retrieval. Products are accompanied by concepts that are difficult to quantify, such as "how stylish is it," and these can change depending on who is asking, or even between Milan and Ottawa. This is why a specialized engine is needed, according to the logic.

In a Day 2 discussion, when Harley Finkelstein directly asked Tobi Lütke why he was so enthusiastic about it, Tobi replied, "Really, really amazing companies will emerge on top of Catalog." The reason was that this foundation previously didn't exist, and anyone trying to build product search had to borrow tools designed for information retrieval. That's why, he said, experts built it from scratch, starting with the very first line of C++.

API keys are prominently placed in the Dev Dashboard, and Shopify has clearly indicated that it has no intention of building products on top of this, wanting partners to fully leverage it. Tobi himself said he gives his agents a budget and APIs to shop, and through that, he has discovered new merchants.

During the keynote, two Catalog-related features were also introduced: Personalization, which reorders search results using signed-in user data, and Commission, where app developers receive a fee when a purchase is made. In terms of "opportunity" for partners, this might be a more direct discussion than UCP.

3. Still, "Stores" Will Not Disappear

On the other hand, it was emphasized that even if AI becomes the entry point, storefronts (store websites) will not disappear. Two pieces of data were presented: AI-driven conversions are 80% higher than organic search, and 50% of AI sessions land on product pages. Even when AI is involved, the site itself is still part of the transaction—the explanation was that appearance, speed, and the narrative being told support the judgment of whether to trust a brand.

AI excels in categories where attributes are standardized and easy to compare, such as electronics, supplements, and watches. Conversely, the challenge they are currently investing in is how to handle items like fashion, apparel, and lifestyle products, where choices aren't made simply by size or color.

4. Sidekick: AI Standing Alongside Merchants

Sidekick, the AI assistant in the Shopify admin panel, was also a major topic. The App Extensions mechanism, which allows integration with external apps, was released in June, shaped in collaboration with 18 partners. As of approximately six weeks after its release, there have already been 500,000 conversations involving operations of these 18 partner apps, and over 160 apps have been integrated. In another session, it was revealed that Sidekick made 467,000 calls to app functions in just two weeks from its release to the end of June.

There were also stories of merchants creating small apps themselves by asking Sidekick. This wasn't a new announcement at DotDev but a feature released at the previous Shopify Editions Winter '26. The content ranged widely from inventory replenishment reports to warranty management, and it was explained that since Shopify barely advertised it, merchants discovered this feature through their interactions with Sidekick.

On the other hand, the design consistently requires human approval before Sidekick can perform operations that modify data. Drafts are created, but actual changes do not occur until the merchant reviews and clicks save. This reaffirmed Shopify's clear strategy of delineating between what AI handles and what humans control. Efforts to enable store operators to confidently leverage AI's power are consistently built in.

5. Even the Tools for Creators are Being Rebuilt for AI

The tools for site builders were also universally re-evaluated for the AI era.

The templating language Liquid, through the process of JSONifying templates to allow no-code editing of any page from the theme editor, had resulted in less readable backend code. However, in response to the era where anyone can edit code with AI, a policy shift was announced to return to a format that is easier for AI to edit. The goal is a declarative interface where one can simply say, "Make this button blue." The new "Base" theme has shrunk collection templates to about 100 lines, representing 93% less code than the current Horizon theme.

Hydrogen, used for headless configurations, is being rebuilt in partnership with Vercel to be framework-agnostic. The interesting part is the backstory: both companies independently arrived at the same conclusion — to open up the core and allow framework choices — about six months prior, and since the differences were mainly in package namespaces, they quickly teamed up.

The analytics platform for app developers is also being revamped. Previously, app developers had to implement analytics themselves, but a vision was presented for closer integration with Shopify's native analytics. Data, once deposited with Shopify, can be reused for reports and automation. The gist was, "The data you send today will be the data an AI agent queries tomorrow."

In the discussion about the future of the web between Tobi Lütke and Vercel's Guillermo Rauch, an agent-operated web was discussed as a given. The conversation revolved around a return to things that both humans and agents can read and modify: file system-based routing, CLIs, small composable programs, and open standards. The retooling, it felt, was preparation for this.

6. Shopify Itself is Powered by AI

The session revealing the behind-the-scenes of Shopify's internal AI agent, River, was also impressive. It was recently featured on Shopify's tech blog and covered in our company's column.

River is a coding assistant that runs on internal Slack, researching code, consulting on design, creating pull requests, and even circulating them for review. The notable feature is that all of this happens in a public space. Anyone can peek into conversations and interject at any point. In five months since its launch on February 13, it has seen approximately 275,000 sessions, used by 90% of employees, and has led to 27,000 merged pull requests (8,000 in June alone). Roughly one in five pull requests created in Shopify's repositories in the last month has been via River. However, the decision to merge is still made by a human.

More than special technology, the fact that it is built upon the accumulation of simple materials like public interactions and work history was insightful. The goal was described as a form of software that observes its internal state and external environment, identifies deficiencies, and rewrites itself—a way of being that begins to behave like a living creature, a phrase that well captured the atmosphere of the day.

It was also discussed as a personal practice. In a Day 2 discussion, Tobi referred to his agent as "clanker" and recommended trying to build an agent like OpenClaw using a service that costs about $5. He said it was messy, but wonderfully messy, similar to the internet when he first fell in love with it.

7. Experiencing AI Hands-On at the Venue

At the exhibition booths, demonstrations that translated the announcements into experiences were popular. "Fit Check," a virtual try-on that allows real-time try-on of everything from full outfits to shoes and hairstyles with just one product photo, is said to work with millions of items in the catalog without needing 3D models. It was explained that the learning of materials and physics was effective: metals shimmered, suede remained matte, sequins caught the light, and fringes and chains moved naturally.

Another "magic mirror" works with just a camera, screen, and PC. It delivers personalized messages and product suggestions to each visitor, and connects to other systems to handle discount code generation and receipt printing on the spot. What started as a prototype with a webcam, monitor, and laptop is now being used by thousands of customers.

There's a follow-up story about this try-on mirror. In the Day 2 discussion, Harley Finkelstein himself revealed that he was so impressed after trying it the day before that he immediately FaceTimed his favorite merchant and told them, "You have to see this." It was striking that the AI, which had been a topic in slides, was now moving right before their eyes and within tangible reach—an experience that the executives themselves shared.

What Remains: Data, Taste, Distribution

These three words concluded Vanessa Lee's keynote. The argument was that if anyone can write code, the value of code itself will decrease. However, the remaining sources of value are unique Data, Taste (sensibility and perspective), and Distribution (the ability to deliver and connect).

For agencies, the question is whether they possess their unique angles and insights. As AI drives efficiency, what will remain as a differentiator? Our company took this conclusion, placed at the end of two days that seemed entirely focused on AI, as a question for ourselves.

The Next Day Was Community Day

On July 23rd, the day after the main DotDev event, we also participated in Brightwork Vol.01, a community event hosted by ShopDev Alliance and Design Packs. Around 200 people gathered for an intimate event consisting of six 15-minute talks, one panel discussion, and plenty of time for networking.

What stood out was that all the presenters, independent app developers and agencies alike, discussed what would remain as a business strength in the age of AI. I believe it was a day that corroborated the answer of Data, Taste, and Distribution, which had been presented in the main conference the day before, from the perspective of practitioners. I will also cover this in a follow-up report.

First, we've conveyed an overview of the event and the atmosphere on site. In upcoming installments, we plan to delve into one theme at a time, such as UCP and Catalog, Sidekick, and changes in storefront creation methods. Stay tuned!