> Published: 2026-06-29 · Category: Product Launch / Update > Source: Cursor Blog — Build from anywhere with Cursor for iOS > Original link: https://cursor.com/blog/ios-mobile-app
What Happened
On June 29, Cursor officially launched its iOS app in public beta, putting the entire IDE into an iPhone application. It is the first truly complete mobile release in the "AI coding tools go mobile" trend of the past two years — not a web wrapper, not a code reader, but a native app that can launch agents, monitor them, and merge PRs directly from a phone.
The public release centers on four capabilities:
1. Launch agents from your phone. Open Cursor iOS, pick a repository, describe a requirement, and the agent gets to work. Like the desktop version, you can choose any frontier model, use voice input, and use slash commands. Unlike desktop, the agent can run in the cloud or on your own computer — even while you're on your way home.
2. "Remote control" mode. When the agent runs on your own machine, you can keep directing it from your phone. Cursor provides a toggle to keep your computer awake — as long as the lid stays open, the agent keeps running from afar.
3. Lock-screen notifications + Live Activities. When an agent finishes, needs confirmation, or is ready to merge a PR, iOS lock-screen widgets push the update to you.
4. Seamless local–cloud handoff. The most notable design decision of this release: an agent running locally can be "handed off" to the cloud mid-task, and vice versa — a cloud agent can be pulled back to local for testing.
Cursor's blog cites three real use cases:
- Getting paged for a production incident and launching an investigation agent from your phone, with the PR ready by the time you reach your desk;
- Responding to a user-reported bug on X by screenshotting, annotating, and handing it to an agent for design fixes;
- Reproducing, diagnosing, and fixing an urgent customer-reported bug entirely from a phone.
- IDEs becoming collaboration platforms (Notion embedding the Cursor SDK was an earlier signal);
- IDEs becoming agent runtimes (Anthropic's Claude Code bypasses the IDE entirely in the terminal);
- IDEs becoming mobile endpoints (today's step).
- No Android version is an obvious gap; Cursor's blog says nothing about an Android timeline.
- Local–cloud agent migration involves state serialization, network interruption recovery, and security permissions — real-world stability needs six months to a year of observation.
- Mobile + long-running agents + keeping personal computers awake = a triple risk of battery, security, and privacy, with no complete privacy policy response from Cursor yet.
Until July 5, running Composer 2.5 on mobile gets a 25% discount — Cursor's usual "product launch + model promotion" combo.
Analysis
Why this matters more than it looks. Cursor has led the AI coding IDE category, but over the past 18 months the IDE category itself has been pulled in three directions:
The iOS app pushes the third direction forward. Cursor isn't building a "code reader" but an "agent launcher + monitor + approver." This differs from Figma bringing design files to mobile: design review is read-heavy, whereas the real mobile pain point for code was never reading code — it's approving changes, monitoring agents, and handling surprises.
The local → cloud → local handoff is the real product innovation. However good iOS autocomplete gets, it can't replace a desktop IDE given keyboard constraints. But "mobile launch + desktop execution" stitches together the phone's anywhere-access with the desktop's horsepower. The scarce asset here isn't iOS coding capability — it's holistic modeling of a developer's full day of work.
What is Composer 2.5? Cursor's own agentic coding model. The promotion hints at two things: Cursor is pushing mobile adoption via "exclusive model + limited-time discount," and mobile inference may be a key battleground for in-house models like Composer 2.5 (since OpenAI and Anthropic mobile experiences aren't optimal).
Why It Matters
1. The AI coding category definition is being rewritten. Within 12 months, nearly every AI coding IDE will be expected to "at least launch agents from a phone." Not nice-to-have — table stakes. 2. Mobile developer workflow becomes a new differentiation battleground. Next challenges: Android? PWA? Linux desktop? Whoever stitches these workflows together first captures the next wave. 3. Local–cloud agent migration is the next engineering challenge for agent runtimes. Cursor is the first to make it a consumer-grade experience — previously only enterprise platforms like Azure AI Foundry and AWS Bedrock tackled similar problems. 4. Vibe coding really is going pocket-sized. When agents can launch from your pocket, the friction of "idea to code" drops toward zero.