Cursor Now Runs on iPad

Cursor, Anysphere’s AI code editor from @cursor, published an official changelog on July 29, 2026 announcing Cursor for iPad on all paid plans. The release deals with a very practical problem: agent work often finishes when you are not at your desk, but review still needs real context. Cursor for iPad is the iPad version of Cursor’s mobile app for creating, reviewing, and merging agent-generated code changes. The takeaway is simple: mobile Cursor is no longer just a place to glance at chats; it now has enough PR surface to make a real review possible.

That matters most when your repo already depends on Cursor rules, AGENTS.md boundaries, cursor skills, or custom agents. The more work you delegate, the more you need the handoff to include diffs, checks, comments, and reviewer state. Cursor’s iPad release is interesting because it moves that handoff onto a larger mobile screen without pretending the iPad is the same thing as a desktop IDE.

See the whole PR, not just the chat

The headline change is the review surface. On iPhone and iPad, Cursor now shows the full PR path: comments, checks, approvals, reviewers, and the ability to ask the agent to resolve comments.

That replaces a shaky habit many agent users know too well: reading the chat summary, trusting that the PR is probably fine, and promising to inspect the rest later. The new surface lets you stay in the review loop at the moment the agent finishes.

A concrete example: an agent updates a billing service and opens a PR. On the old mobile pattern, you might only see the chat transcript and the final PR link. With the new review screen, you can read a failing check, add the payments maintainer as a reviewer, ask the agent to address a comment about idempotency, and wait for the updated PR before approving.

The trap is treating full PR access as full certainty. A green check and a tidy diff still do not prove the change is safe. For risky code paths, the iPad review should narrow the queue, not remove the need for deeper desktop inspection.

Use the bigger screen for parallel agent work

Cursor says the iPad layout is rebuilt around the extra space. Sidebar chats stay pinned, so you can watch several agents at once. Split screen keeps a review open beside a chat, and file diffs render in full.

That is the part developers will feel first. Agent work is often parallel by nature: one chat fixes a flaky test, another drafts a migration, another updates docs. On a phone, that feels like jumping between cards. On an iPad, the product can show more of the actual work at the same time.

Markup also gets more room. You can attach a screenshot, tap to drop a comment at a specific point, or draw directly on the image with Apple Pencil. That is useful for UI bugs where the best review comment is not a paragraph, but a circle around the thing that moved three pixels.

The limit is still important. The iPad is better for monitoring, review, markup, and routing. It is not a magic replacement for deep debugging, heavy local repro, or long architecture edits where you need several files, logs, traces, and tests in view.

Let Inbox become the handoff receipt

Cursor also added an Inbox for iPhone and iPad. It shows what is in progress, what needs your attention, and which PRs are in review.

That is small in the way good workflow changes are small. The agent does not just produce output; it produces states. Waiting, blocked, ready for review, needs a human, merged. Inbox gives those states a place to land.

The changelog also mentions multi-PR sessions. When one chat creates multiple PRs, you can now open every one of them, not only the last. That matters for a common split: one PR changes code, another updates tests, another patches docs or generated types.

Cursor added Bitbucket and Azure DevOps source control support too, plus team switching inside the app. The practical read is that mobile agent review is moving beyond a narrow GitHub-only side path. The trap is letting Inbox become another notification pile; it works best when a PR has a clear next human action.

Try it in one repo before changing habits

Start with one repo and one low-risk agent PR. Good candidates are docs fixes, test-only changes, small UI copy changes, or refactors with strong test coverage.

Before you try it, make the review boundary explicit. If the repo has Cursor rules or an AGENTS.md file, add a short mobile review note that says what can be approved from iPad and what must wait for desktop. This is also a good moment to revisit the related training topic if your workflow depends on Cursor subagents and skills.

Here is a small Cursor rule stub you can adapt:

---
description: Use when reviewing agent-created PRs from Cursor mobile
globs:
  - **/*
---

Mobile PR review boundary:
- Approve from iPad only when the diff is small, tests are visible, and the changed area has an owner or clear rule.
- Do not approve database migrations, auth changes, billing changes, or permission changes from mobile alone.
- When a screenshot is attached, verify the linked code diff and not only the image.
- If the agent resolves a comment, reopen the changed files before approving.
- Leave a review note with what was checked: diff, comments, checks, and risky files.

If you package repo habits into plugins, the same boundary can later live beside commands, skills, and agent definitions. For a concrete packaging pattern, see Create a Cursor Plugin You Can Install.

What to try first

Copy this checklist into one repo issue or PR template and use it on a single agent-created PR.

• Pick a small PR created by a Cursor agent.
• Open it from Cursor Inbox on iPad.
• Read the full diff before reading the agent’s summary.
• Check comments, checks, approvals, and requested reviewers.
• Ask the agent to resolve one review comment.
• Reopen the changed file after the agent responds.
• Add a short review note: what you checked, what you skipped, and whether desktop follow-up is needed.
• Merge only if the repo rule allows that class of change from mobile.

The useful measurement is not whether mobile review feels faster. It is whether the merged PR needed avoidable follow-up later. If the answer is yes, tighten the boundary before approving more from iPad.

Common questions

• Is Cursor for iPad available on the free plan?

  No, the July 29, 2026 changelog says Cursor for iPad is available on all paid plans. It does not describe free-plan availability in that announcement, so treat paid-plan access as the official claim unless Cursor updates the changelog or pricing docs later.

• Can I review an entire PR on iPhone too?

  Yes, the full review surface applies to both iPhone and iPad. Cursor says the mobile review screen now covers comments, checks, approvals, reviewer changes, and agent resolution prompts, while the iPad gets the larger rebuilt layout for pinned chats, split screen, and fuller diffs.

• Does Cursor on iPad replace desktop Cursor?

  No, it is better understood as a mobile review and agent-control surface. The release makes create, review, and merge possible from iPad, but deep debugging, broad refactors, local reproduction, and high-risk architecture changes still deserve a desktop workflow with more context.

• How do Cursor rules help with iPad review?

  Cursor rules help by making the review boundary explicit before you are making a merge decision on a smaller device. A short .mdc rule can say which files, risk areas, and PR types are safe to approve from mobile, and which ones require desktop follow-up.

• What changed for chats that create multiple PRs?

  Cursor now supports multi-PR sessions on mobile, so you can open every PR from a chat instead of only the last one. That matters when an agent splits work across code, tests, docs, or generated files and each PR needs a separate review decision.

Best ways to use this research

• Best for: understanding what Cursor for iPad changed in the July 29, 2026 official release, especially around mobile PR review and agent handoff.
• Best first artifact: a one-repo mobile review checklist that says which agent PRs can be approved from iPad and which must wait.
• Best comparison angle: compare chat-only mobile review against full PR review: diffs, checks, comments, reviewers, approvals, and post-comment agent fixes.
• Best measurement: track whether mobile-approved PRs need avoidable follow-up. If they do, the boundary is too loose, not the reviewer too slow.

Further reading

• @cursor — source
• Cursor — Agent
• Cursor — Rules

Start with one small PR

Open one low-risk agent PR on iPad, review the full diff and checks, then write down what you would not approve from mobile next time. That boundary is the real feature you get to keep.

One methodology lens

One useful way to read this through our methodology is the Plan step: delegate first-pass decomposition and dependency mapping, review the sequencing and assumptions, and keep ownership of scope and priorities. If that split is still fuzzy, the workflow usually is too.