AI review and AI agents
Both are part of the unlimited plan, and both are optional. Neither changes what the scan finds: the score, the blockers and every check are the same with or without them.
AI review, only if you ask for it
AI-assisted review is part of the unlimited plan, and it is off until you switch it on: the switch is on the scan screen, it starts off for everyone, and the answer stays yours to change.
What it does
- Clearer permission wording. It reads the purpose strings already in your Info.plist, says which are vague or still placeholder text, and suggests wording that states the reason a reviewer is looking for.
- A finding in plain language. The same finding, explained for somebody meeting the rule for the first time. It is appended under the deterministic explanation, never in place of it, so you can read both.
- Its own uncertainty, in writing. Where the evidence did not settle something, it has to say so, and that line is printed with the scan limitations rather than smoothed away into a confident sentence.
Four things it is never allowed to do
Each of these is a property of the code rather than of the prompt. A model told not to do something and a model with nowhere to put it are different promises, and only the second one is worth printing.
- Add a finding. Output naming a rule that did not run matches nothing and is discarded.
- Remove one. It is handed findings to explain, never the list to decide.
- Change a severity. Its output has no severity field, and your score is worked out before it is called.
- See your project. It is sent a summary (finding titles and explanations, evidence notes, your permission strings, the bundle identifier and SDK names), never your files, your tree or the evidence excerpts.
Your score, your blockers and your checks are identical either way. Say no and the scan is exactly the one every deployment without a key runs. What the summary contains, and who receives it, is set out on where your code goes and in the privacy policy.
Check your app from your AI agent
Connect ReviewAhead to Claude Code, Codex or any agent that supports MCP. Ask it whether your app is ready and it uploads your project, reads the report and fixes what it can in your code, then asks you what only you can answer.
claude mcp add --transport http reviewahead https://app.reviewahead.dev/mcpThen run /mcp in Claude Code, pick reviewahead and sign in. Codex and other MCP clients take the same address: https://app.reviewahead.dev/mcp
Use reviewahead to check whether my app is ready for the App Store.
Good to know
- It is part of the unlimited plan. You can connect without it, but your agent will be told to subscribe before it can scan.
- Uploading needs an agent that can run commands on your computer, such as Claude Code or Codex. It zips your project and uploads it through a link that works once, for fifteen minutes.
- Scans from your agent appear in the same history as scans on the website.
- Your agent receives your reports only when you have connected it and it asks for them. What it does with them, including sending them to the AI provider behind it, is between you and that provider.
- You can disconnect at any time in your agent’s settings.