Cursor for iOS Guide 2026: Launch and Manage AI Agents from Your Phone

Cursor for iOS, launched June 29, 2026 in public beta, lets you launch AI coding agents from your iPhone and manage both cloud-hosted and local desktop agents on the go. The app supports voice input, slash commands, and frontier model selection — and with 75% off Composer 2.5 runs through July 5, the timing is designed to get developers hooked early. Cursor crossed $2B ARR in early 2026 with over 1 million daily active users, and the iOS move turns your phone into a remote control for the fastest-growing AI IDE on the market. Here is exactly how to set it up and make it useful.

What Is Cursor for iOS?

Cursor for iOS is a native iPhone application that extends the Cursor AI code editor into a mobile agent management platform. Unlike the desktop IDE, the iOS app does not let you edit files directly — instead, it gives you a dashboard to launch cloud-hosted agents, connect to agents running on your local desktop machine via Remote Control, and review code diffs, artifacts, and PR status from your phone. Cursor has over 1 million daily active users and more than half of Fortune 500 companies have developers using the platform, according to the Anysphere Series C announcement. The iOS app runs on all paid Cursor plans (Pro at $20/month, Pro Plus at $40/month, and Ultra at $200/month) and requires an existing Cursor account. The core value proposition is unblocking yourself anywhere: approve a diff at lunch, restart a stuck agent from the airport, or review a PR while commuting. Cursor crossed $2 billion ARR in early 2026, and the iOS launch is a direct response to developer demand for mobile agent management rather than mobile code editing.

What You Need Before You Start

Before installing Cursor for iOS, you need an active Cursor paid subscription — the iOS app requires either Pro ($20/month), Pro Plus ($40/month), or Ultra ($200/month). The free tier does not include mobile access. You also need a Cursor account with at least one project repository connected via GitHub, GitLab, or Bitbucket. The iOS app uses OAuth 2.0 to authenticate with your existing Cursor account, so ensure your Cursor desktop client is up to date (version 0.45 or later). If you plan to use Remote Control to connect to agents on your local machine, your desktop Cursor must be running and signed into the same account on a network that allows outbound connections — the app does not support LAN-only setups. Cloud agents, which run on Cursor’s infrastructure, work immediately without any desktop dependency. According to JetBrains’ January 2026 survey, Cursor holds 18% market share among AI coding tools (second behind GitHub Copilot at 29%), meaning a significant portion of developers can already use the iOS app without switching tools. The 75% discount on Composer 2.5 runs through July 5, 2026, so activating cloud agent runs during the promotion window cuts your per-task cost substantially.

How to Install and Set Up Cursor for iOS

Installing Cursor for iOS starts by downloading the app from the iOS App Store — search for “Cursor AI” and look for the public beta release from Anysphere. The app requires iOS 17.0 or later and an iPhone with at least 6GB of RAM for smooth agent status rendering. After installation, you authenticate via OAuth 2.0 with your Cursor account credentials, which grants the mobile app access to your projects, agent history, and connected repositories. You do not need to reconfigure API keys or model preferences — the app pulls your existing desktop settings from Cursor’s cloud sync. The setup process takes under two minutes: download, authenticate, and the main dashboard loads with your recent agent sessions and project list. Cursor crossed $2 billion ARR faster than any other AI developer tool, and the iOS onboarding reflects that engineering investment — zero configuration required beyond the initial login. For teams using Cursor Ultra ($200/month), the iOS app also surfaces organization-level agent logs and pending review requests, though admin controls remain desktop-only. After login, test the connection by tapping any recent agent session — the app loads the status, recent logs, and any pending approval requests.

Launching Your First Cloud Agent from Your Phone

Launching a cloud agent from Cursor for iOS requires tapping the new agent button in the bottom-right corner of the dashboard, selecting a project repository, and typing your task using either voice input or the keyboard. The app supports slash commands — /fix, /test, /refactor, and /explain — that prepopulate agent instructions with proven prompt templates. You can also select the underlying AI model: Cursor’s in-house Composer-1 or Sonic, Anthropic Claude, OpenAI GPT models, Google Gemini, or xAI Grok. Cursor holds 18% of the AI coding tool market and serves over 1 million daily active users, which means cloud agent availability depends on demand during peak hours. The cloud agent runs on Cursor’s infrastructure, not your phone, so you can lock the screen or switch apps while it works — push notifications will alert you when the agent completes or requires input. A typical cloud agent task — such as “add input validation to the login form” — completes in 30 to 90 seconds depending on model and complexity. The 75% discount on Composer 2.5 runs through July 5, making each cloud agent run roughly $0.08 at the Pro tier instead of $0.32. After the agent finishes, review the diff inline, request changes, or approve and commit directly from the notification.

Using Voice Input for Agent Commands

Voice input on Cursor for iOS converts speech to text using Apple’s on-device speech recognition and sends the transcribed prompt to the selected agent. This is useful for dictating bug descriptions while walking, describing refactoring intent hands-free, or capturing thoughts before they fade. The voice system supports natural language — saying “add error handling to the payment webhook and test it with Stripe’s sandbox” generates the correct agent instructions without manual formatting.

Using Remote Control for Local Machine Agents

Remote Control lets you connect to agents running on your Cursor desktop client from the iOS app, giving you visibility and control over locally executing tasks. Enable it from the app’s settings by selecting your desktop machine from the device list — both devices must be signed into the same Cursor account, and your desktop must have Remote Control toggled on in Cursor’s preferences. Once connected, the iOS app streams live agent logs, terminal output, and file diffs from your desktop agent to your phone in real time. You can approve or reject file changes, cancel stuck agents, and send follow-up instructions without touching your keyboard. More than half of Fortune 500 companies use Cursor, and enterprise teams use Remote Control primarily for on-call incident response — triaging production issues while away from their desk. The connection uses end-to-end encryption, and agents running locally never expose your source code to Cursor’s cloud infrastructure. One practical pattern is starting a long refactoring task on desktop, walking to a meeting, and monitoring progress on your phone — if the agent gets stuck on a permission prompt, you can approve it remotely and let the agent continue.

Security Considerations for Remote Control

Remote Control requires biometric authentication (Face ID or Touch ID) to activate the connection, and the session expires automatically after 15 minutes of inactivity. You can also revoke access from any device at any time. Cursor uses OAuth 2.0 with short-lived tokens for mobile-to-desktop communication, and no API keys or credentials are stored on the iOS device itself.

Staying in the Loop — Live Activities and Push Notifications

Cursor for iOS uses Apple’s Live Activities API to put agent status directly on your iPhone’s lock screen and Dynamic Island, so you can see agent progress without unlocking the phone. When a cloud agent finishes, finds an error, or requires approval, the app sends a push notification with the agent summary and a quick-action button to approve, dismiss, or request changes. For locally running agents connected via Remote Control, notifications mirror the desktop Cursor notification system — build failures, test results, and review requests all appear on your phone with the same priority levels you set on desktop. Cursor crossed $2B ARR by shipping features developers actually use, and the notification system was built based on user research showing that the top mobile request was “don’t make me poll the app — tell me when something needs me.” You can customize notification channels per project: mute non-critical agent activity on your personal project while allowing all notifications for your team’s production repository. Live Activities also display estimated time remaining for cloud agents, updated every five seconds, so you know whether to wait or check back later.

Reviewing Work — Diffs, Artifacts, and PR Merging

The Cursor for iOS review interface presents code diffs in a mobile-optimized format with syntax highlighting, side-by-side or unified view options, and tap-to-expand file sections. Each diff shows the changed lines, the agent’s reasoning summary, and a confidence score for the change. You can leave inline comments, request a revised approach, or approve and merge directly from the app. For cloud agents, approved changes commit to your repository via the connected GitHub, GitLab, or Bitbucket integration — you do not need to return to desktop to land the work. Artifacts — previews of generated UI components, documentation, or configuration files — render inline in the app using WebKit, with zoom and scroll support. 90% of Salesforce’s engineering organization uses Cursor daily, and they report that mobile review has cut their PR cycle time by roughly 40% for single-agent tasks because reviewers can approve changes during downtime that would previously have waited until the next desk session. The diff view also supports applying partial changes — you can approve three of five file modifications and reject the remaining two, sending targeted revision instructions back to the agent.

Handoff Between Local and Cloud Environments

Cursor for iOS enables seamless handoff between cloud agents and local desktop agents through a shared session history. Start a task as a cloud agent from your phone, and if you decide it needs deeper local access — for example, to run database migrations or access environment variables — you can transfer the session to your desktop Cursor with a single tap. The handoff preserves the full agent conversation, all file diffs generated so far, and the agent’s working state. Conversely, if you start a long-running refactoring task on desktop and need to leave, you can convert it to a cloud agent that continues executing on Cursor’s infrastructure while you monitor from your phone. Cursor is the fastest SaaS company to reach $100M ARR, and this handoff capability removes the biggest adoption barrier for mobile agent management: losing context when switching devices. The handoff works bidirectionally for Pro Plus and Ultra plans; Pro plans can hand off from phone to desktop but not from desktop to cloud. The transferred session appears in the iOS app’s recent agents list with a Cloud or Local badge so you always know where each agent is executing.

Real-World Workflows — Incident Response, Bug Fixes, Design Feedback

Cursor for iOS supports three primary real-world workflows that justify the mobile investment beyond novelty. Incident response: when your monitoring system alerts you to a production error, open Cursor iOS, select the affected repository, and launch a cloud agent with a slash command like /fix "500 error on checkout endpoint — check Stripe webhook handler for timeout regression". The agent investigates, proposes a fix, and you approve from the notification — all while walking back to your desk. Bug reproduction: a teammate reports a UI bug with a screenshot. Use voice input to describe the issue, launch a cloud agent to write a failing test that reproduces the condition, review the test on your phone, and merge it to your branch. Design feedback: during a design review, a teammate points out a spacing issue. Launch an agent with /refactor "adjust Card component padding from 16px to 24px on mobile breakpoint", approve the diff on your phone, and the PR updates before the meeting ends. Cursor has over 1 million daily active users, and these three patterns account for roughly 65% of iOS app usage according to Cursor’s internal analytics shared at the launch event. The key is that none of these require mobile code editing — they use agents to do the work while you make decisions.

Cursor for iOS vs Competitors — Copilot, Claude Code, Windsurf

GitHub Copilot Mobile offers a chat-only experience that answers questions and generates short code snippets, but it cannot launch agents, review full PR diffs, or control a desktop session. Claude Code runs entirely in the terminal with no mobile client — you can SSH into a remote machine from your phone, but you lose visual diff review and notification-driven workflows. Windsurf has not released a mobile app as of July 2026. Cursor for iOS is the only tool that treats the phone as a remote agent operations center rather than a secondary chat surface. Cursor holds 18% market share in JetBrains’ January 2026 survey, and the iOS feature set suggests Anysphere is targeting the 59% of developers who already run three or more AI coding tools in parallel — making Cursor the hub for agent management across devices.

Tips and Best Practices for Mobile Agent Management

Use slash commands instead of free-form prompts for faster agent launches — /fix, /test, /refactor, and /explain produce better results than typing full paragraphs on a phone keyboard. Keep cloud agents for tasks under five minutes of compute time; anything longer should run locally or use Remote Control to avoid per-run costs eating into your quota. Enable push notifications only for production repositories to avoid alert fatigue — personal projects can be checked on demand. The 75% discount on Composer 2.5 runs through July 5, 2026, so batch your cloud agent work into this window if you are on a Pro plan. For Remote Control sessions, make sure your desktop Cursor has “Auto-approve safe changes” enabled in settings — this reduces the number of notifications you need to act on from your phone. If you work on a team using Cursor Ultra, assign the iOS app a dedicated notification profile that separates your on-call repository from your daily work. Cursor crossed $2B ARR in part by listening to power users, and the iOS features — Live Activities, voice input, handoff — all came from direct developer requests in Cursor’s public feedback board. When reviewing diffs on mobile, use the unified view for single-file changes and side-by-side for multi-file reviews — unified renders faster on cellular connections.

Pricing and Plan Considerations

Cursor for iOS is included in all paid plans — there is no separate mobile subscription. The Pro plan ($20/month) gives you 500 Composer 2.5 agent runs per month plus 500 fast premium model requests. The Pro Plus plan ($40/month) doubles those limits to 1,000 runs and 1,000 fast requests, and adds bidirectional cloud-to-desktop handoff. The Ultra plan ($200/month) includes unlimited agent runs, team management features, organization agent logs, and priority infrastructure access. The 75% discount on Composer 2.5 runs — active through July 5, 2026 — reduces per-run costs significantly: at Pro tier, each run costs roughly $0.08 during the promotion instead of $0.32 after. Cursor is the fastest SaaS company to reach $100M ARR, and the pricing strategy for iOS suggests Anysphere uses the mobile app as an upsell funnel — the free desktop tier cannot use iOS, and the mobile experience intentionally showcases cloud agent capabilities that require paid runs. If you primarily use cloud agents from iOS, the Pro Plus plan is the sweet spot because handoff lets you start on phone and finish on desktop without losing the context. Enterprise teams on Ultra should budget for increased cloud agent usage as developers adopt mobile workflows — Cursor’s internal data shows iOS users consume 22% more agent runs in their first week than desktop-only users.

Frequently Asked Questions

Cursor for iOS launched June 29, 2026 as a public beta on all paid plans, and developers evaluating the app consistently ask the same five questions about capabilities, device support, costs, offline access, and security. These answers reflect the current state of the iOS app — Anysphere has not announced a roadmap for future mobile features as of July 2026, but based on the company’s $2B ARR trajectory and 1 million DAU base, the iOS app will likely see rapid iteration based on user feedback collected through the public beta. Cursor crossed $2 billion annualized revenue in early 2026 with over 1 million daily active users and more than half of Fortune 500 companies having developers on the platform, and the iOS public beta represents the company’s first major platform expansion beyond desktop. The answers below cover the most frequently asked questions that arose during the first week of the public beta release.

Can I edit code directly in Cursor for iOS?

No, Cursor for iOS does not include a code editor. The app is designed for launching and managing AI agents, reviewing diffs, and approving changes — not for writing code directly on your phone. All code changes are executed by agents running in the cloud or on your local desktop machine.

Does Cursor for iOS work on iPad?

Cursor for iOS runs on iPhone only at launch. The app does not have an iPad-optimized version, though it runs in compatibility mode on iPad at iPhone resolution. Anysphere has not announced iPad support publicly as of July 2026.

How much does each cloud agent run cost?

At the Pro tier ($20/month), each Composer 2.5 run costs approximately $0.32 at standard rates. Through July 5, 2026, a 75% discount reduces this to roughly $0.08 per run. Pro Plus ($40/month) includes twice the runs at the same per-run rate, and Ultra ($200/month) includes unlimited runs.

Can I use Cursor for iOS offline?

No, Cursor for iOS requires an active internet connection. Cloud agents need connectivity to run on Cursor’s infrastructure, and Remote Control requires network access to communicate with your desktop machine. Live Activities and cached session history are available briefly without connection, but agent operations require connectivity.

Is Cursor for iOS secure for enterprise use?

Yes, the app uses OAuth 2.0 authentication, biometric unlock for Remote Control sessions, and end-to-end encryption for agent log streaming. No source code is stored on the device. Enterprise teams on Ultra can enforce device-level security policies, including mandatory Face ID and remote session revocation.