Shopify has built Remote Dev, a remote environment where AI agents build, test, and maintain iOS and Android apps on a fleet of remote Macs and Linux virtual machines instead of developers’ laptops.
Shopify moved mobile AI agent work off developer laptops and onto remote Macs and Linux VMs, a system it calls Remote Dev. Agents build, test, and open pull requests in parallel environments, while engineers reach the machines over SSH.
Why did laptops max out?
Shopify’s mobile teams ran AI agent workflows—the kind that edit code, run builds, and retry failures—directly on their own machines. The company reports that three or four workflows running at once were enough to max out a laptop, on top of what compilers, simulators, and emulators already consume.
There was a more mundane failure mode too. The company said that “if someone accidentally closed a laptop lid, the work stopped.”
Shopify had already solved this for web and backend development with River, a Linux-based environment where its AI agent operates. That solution couldn’t carry over to iOS: building and testing iPhone apps requires Xcode, which only runs on macOS.
How does Remote Dev work?
The core is a fleet of remote Macs, each hosting Shopify’s development environment and iOS simulators, reachable over SSH—the remote-access protocol that lets developers and agents log in from anywhere. Android emulators run on separate Linux virtual machines, so a single SSH session can build and test both platforms.
Each agent session gets its own isolated environment. One agent can debug an Android problem while another works on an iOS feature and a third tests a pull request. Tasks can also arrive through a Slack channel backed by River.
The agents can change code, build the app, run tests, drive the simulators, collect screenshots, and open a pull request. When a step fails, an agent reads the logs, fixes the problem, and tries again while the engineer does something else.
Overnight automated runs look for bugs, implement fixes, and submit pull requests; agents can also verify a build by using the app's interface, reproducing a bug, fixing it, and repeating the same steps to confirm.
How widely are agents used?
The clearest sign of adoption is a single number Shopify reports: River’s share of merged pull requests rose from 6% to 61% after the company pushed internal adoption through demonstrations and internal storytelling.
Personnel changes rarely come with usage charts, so this figure is the main quantitative evidence the company offers that mobile engineers actually shifted their work onto agent-run environments.
Is it really autonomous?
A LinkedIn commenter who leads software infrastructure and founded SpendMe argued that Remote Dev’s main advantage is its integration with Xcode and the iOS Simulator—and that the end result still depends on a human at a Mac.
A PR you still have to pull and verify on a Mac isn’t really autonomous.
the LinkedIn commenter, writing on Shopify’s article
He added that the write-up leaves open how builds avoid queuing when several agents request a Mac at the same time.
A second commenter, a solutions architect, focused on test integrity. If agents run the simulators and the tests, he warned, results may not be reliable.
Give [the agent] a red test and enough retries and it’ll sometimes rewrite the test until it passes.
the solutions architect, whose proposed fix is to bar runner agents from modifying the tests they execute
What should teams copying this plan for?
Teams following Shopify’s design need to plan for Mac capacity, because iOS builds depend on it, and for scheduling, so agents don’t stall waiting for the same machine. Deciding early which agents may edit tests and which may only run them addresses the test-integrity concern.
What does Shopify plan to add?
Shopify says it intends to extend the platform with continuous application health monitoring, automated crash detection and fixes, and automatic pull request generation for problems it finds. No timelines have been announced.