Cursor for iPad: Setup, PR Review and Limits

The iPad app uses the same agent backend as cursor.com/agents and the desktop Agents WindowCursor's surface listing your agent runs; open a run to read its diff and search transcripts from the command palette. Press Enter for the full definition.. An agent started on the tablet appears on the other surfaces, and an existing cloud agent appears in the mobile inbox. From the iPad you choose a repository and branch, send a task, follow the live transcript and reply while the agent runs. You can review the result later without keeping the iPad awake.

The July 29 release rebuilt the layout for the larger screen: chats stay pinned in a sidebar, a review can sit beside the active conversation, and diffs use the full width. With several cloud agents running at once, those changes turn the tablet from a notification endpoint into a workable review station.

Sign in with the account you use on desktop or the web, so the iPad sees the same agents and repositories. The agent runs in its selected environment, so the tablet needs no local checkout or terminal session of its own.

1. 1Download Cursor from the App Store on an iPad running iPadOS 26.0 or later.
2. 2Sign in. If your organization requires SSOSingle Sign-On. One company login (usually via SAML or OIDC) instead of a separate password per tool. Press Enter for the full definition., you sign in through it first.
3. 3Choose the repository, branch, model and available worker for the run.
4. 4Describe a bounded task, and include the check that should prove it is done.
5. 5Follow the run in the sidebar, then open its pull request or diff for review.

Make the first task one whose result you can review from a diff.

The cloud environment must already contain what the project needs. Secrets, environment setup and source control connections live on the web, not in the iPad app. If the first run cannot install or test the project, fix that setup at cursor.com before you retry. Prompting again from the tablet will not supply a missing credential or build service.

The mobile review covers the whole pull request. Diffs, commits, deployments, comments, checks and approvals all open on the tablet, and you can add or change reviewers. From the same screen, ask the agent to resolve a comment or fix a failed check. Cursor also documents squash merge, mark ready, update branch, toggle auto-merge, publish and close.

Use split screen to keep the review beside the chat. That pairing is what makes the iPad useful for review: a suspicious hunk becomes a precise follow-up without losing the conversation around it. When one chat produces several pull requests, multi-PR sessions let you open each one instead of only the latest.

A merge button does not replace repository policy. Required approvals, protected branches and failing checks can still block the merge. Before you merge from the smaller screen, apply your desktop standard: read the actual diff, confirm the automated checks, and resolve the review state.

Attach a screenshot, then tap the exact point that needs attention or draw on the image with Apple Pencil. Marks are easier to place on the iPad's larger canvas than on a phone. The app also takes photos, camera captures, files and voice dictation, so you can describe a visual bug while its location stays attached to the prompt.

Markup tells the agent where to look and nothing more. A circle can show where a button overlaps, but it does not state the expected responsive behavior, keyboard state or test condition. Pair each mark with one sentence that describes the result you want. When visual correctness matters, ask for a screenshot or a deployed artifact after the change.

The app is not a full IDE or an administration console. It shows changed files in a diff view and has no editor, terminal or file browser. Environment secrets, MCPModel Context Protocol. A standard that lets an AI agent pull in context from outside the repo, like Jira tickets or internal docs. Press Enter for the full definition. server setup, source control connections, rules, skills, automations, admin settings, billing and usage stay on the web. On mobile you can pick an existing MCP server for a run, but you cannot add or manage one.

Cursor for iOS relies on Cloud AgentsAgents that run in a Cursor-managed virtual machine, check out the repo, do the work and open a pull request, then shut down, with no load on your laptop. Press Enter for the full definition., so it needs cloud data storage. Accounts on Legacy Privacy ModeCursor's setting that guarantees code data is not used for training by Cursor or its model providers, and that an admin can enforce org-wide; data-retention terms are a separate, contractual layer. Press Enter for the full definition. must switch to Privacy Mode before they can start mobile agents, and Cursor says that switch cannot be reversed. Cursor does not train on your code. Even so, read the Cloud Agent storage and retention terms before you enable the app for sensitive repositories.

Check that the screen is current. The interface is cache-first: it opens the inbox and conversations from local data before the network catches up. That helps on a weak connection, but an old screen does not prove the remote run is unchanged. Let the app sync before you merge, close or redirect an agent whose current state matters.

Then check the workspace. You can switch between the teams you belong to inside the app, so it is easy to start work under the wrong one.

Confirm the active team, repository and branch before you start a task.