ios-ship-doctor
Enables AI assistants to diagnose and fix iOS app issues that cause App Store rejections, such as missing privacy manifests, permission strings, and test credentials, before submission.
README
<div align="center">
🩺 iOS Ship Doctor
Catch App Store rejections before you hit Submit.
An MCP server that lets your AI assistant — Claude Code, Gemini CLI, Codex, Cursor, Copilot, Windsurf, Zed — diagnose why your iOS app will be rejected — privacy manifests, missing permission strings, unlisted SDKs, leftover test credentials — and then fix them. When Apple does reject you, it pulls the rejection and maps it to a plain-English fix.
<br/>
<img src="./assets/demo.svg" alt="Ship Doctor preflight output" width="720"/>
</div>
Why
Apple's review pipeline silently rejects builds for things you can't see in Xcode:
- A required-reason API used in code but not declared in
PrivacyInfo.xcprivacy - A permission used with no
NS…UsageDescriptionstring (instant crash + reject) - A third-party SDK on Apple's list shipping without its own privacy manifest
- A placeholder test credential (like Google's sample AdMob ID) left in
Info.plist
Each one costs you a submission cycle — hours or days of round-trips. Ship Doctor checks all of them in seconds and tells you exactly what to change.
Existing App Store MCPs are thin API wrappers that submit your app. Ship Doctor is the one that tells you why it won't pass first.
Quickstart
No install needed — npx fetches and runs it:
npx -y ios-ship-doctor-mcp preflight /path/to/your/app
Then wire it into whichever assistant you use. This is a standard stdio MCP server — it is not Claude-specific and works with any MCP client. Let it print the exact config for yours:
npx -y ios-ship-doctor-mcp config # every client
npx -y ios-ship-doctor-mcp config gemini # just one
<details> <summary>Or clone and build from source</summary>
git clone https://github.com/menansali/ios-ship-doctor.git
cd ios-ship-doctor
npm install && npm run build
npm link # exposes the `ios-ship-doctor-mcp` command
</details>
Supported out of the box: claude, gemini, codex, cursor, vscode, windsurf, zed, generic.
| Client | Config file | Root key |
|---|---|---|
| Claude Code | claude mcp add ios-ship-doctor -- npx -y ios-ship-doctor-mcp |
mcpServers |
| Gemini CLI | ~/.gemini/settings.json (or gemini mcp add …) |
mcpServers |
| OpenAI Codex CLI | ~/.codex/config.toml |
[mcp_servers.…] ⚠️ underscore |
| Cursor | ~/.cursor/mcp.json or .cursor/mcp.json |
mcpServers |
| VS Code (Copilot) | .vscode/mcp.json |
servers ⚠️ not mcpServers |
| Windsurf | ~/.codeium/windsurf/mcp_config.json |
mcpServers |
| Zed | Zed settings.json |
context_servers ⚠️ command.path |
| Anything else | — | mcpServers |
The two footguns the generator handles for you: Codex silently ignores mcp-servers (it must be mcp_servers), and the printed node path avoids version-pinned Homebrew/nvm paths that break on the next upgrade.
Then just ask:
Is
/path/to/my-appready to submit to the App Store?
Example
🩺 Ship Doctor preflight — /Users/you/Desktop/MyApp
VERDICT: NOT READY — 1 issue that commonly causes rejection.
(1 error, 1 warning, 6 passed)
❌ [credential-traps] Google AdMob TEST application ID in Info.plist
Found placeholder/test value "ca-app-pub-3940256099942544".
↳ MyApp/Info.plist
💡 Replace GADApplicationIdentifier with your real ca-app-pub-… ID.
⚠️ [export-compliance] Missing ITSAppUsesNonExemptEncryption in Info.plist
Without this key, App Store Connect prompts about encryption on every submission.
đź’ˇ Add <key>ITSAppUsesNonExemptEncryption</key><false/> if you only use HTTPS/TLS.
âś… [privacy-manifest] Privacy manifest consistent with API usage
âś… [usage-descriptions] All permission strings present
âś… [dependencies] SDK privacy manifests OK
âś… [app-icon] App icon asset found
Tools
Preflight — local, no credentials
| Tool | What it catches |
|---|---|
preflight |
Runs every check below → single READY / NOT READY verdict |
scan_privacy_manifest |
Required-reason APIs used but undeclared, invalid reason codes, missing manifest |
check_usage_descriptions |
Camera / location / photos / mic / tracking used with no NS…UsageDescription |
audit_dependencies |
Apple-listed SDKs shipping without a PrivacyInfo.xcprivacy |
check_credential_traps |
Placeholder / public test credentials in Info.plist |
check_legal_links |
Missing Privacy Policy / Terms of Use (EULA) links on a paywall + in the App Store description (3.1.2) |
check_account_requirements |
No demo account for App Review (2.1), missing in-app account deletion (5.1.1(v)), social login without Sign in with Apple (4.8) |
check_external_payments |
Stripe/PayPal/Paddle for digital content with no StoreKit (3.1.1) |
check_background_modes |
UIBackgroundModes entries the app never actually implements (2.5.4) |
check_placeholder_content |
Lorem ipsum, Stripe test keys, YOUR_API_KEY, example.com dead links, template app names (2.1) |
generate_privacy_manifest |
Writes a valid PrivacyInfo.xcprivacy covering every detected API |
autofix |
Applies the safe fixes automatically; reports the rest as manual follow-up |
preflight also checks export compliance, App Transport Security, app-icon presence, banned APIs (UIWebView), launch screen, version/build sanity, and deployment target. Dependency scanning covers both CocoaPods and Swift Package Manager.
Project layouts understood: classic Info.plist projects, modern projects with GENERATE_INFOPLIST_FILE and no plist file (keys read from INFOPLIST_KEY_* build settings), ios/ subdirectories, and monorepos where the .xcodeproj sits a level or two down.
What this does not check
Ship Doctor reads project files. It cannot see the things most rejections are actually about:
- Crashes and incomplete features (2.1) — the most common rejection; needs the app running
- Design quality (4.0) and spam/duplicate (4.3) — human judgement
- Metadata accuracy (2.3) — whether screenshots match the real app
- Privacy label accuracy (5.1.1) — whether declared data collection matches SDK behaviour
- Anything a dependency does; source scanning is first-party only
A clean run means "no automated check fired", not "this will be approved."
Rejection recovery — needs an App Store Connect API key
| Tool | What it does |
|---|---|
asc_list_apps |
Lists your apps (id, name, bundle id) |
asc_get_rejections |
Pulls recent rejections, maps Review Guideline numbers → summaries + fixes |
asc_check_submission |
The metadata half of preflight: demo credentials actually filled in, review notes, required screenshot sets |
explain_guideline |
Explains any Review Guideline number (offline, no key needed) |
<details> <summary><b>Enabling rejection recovery</b></summary>
Create a key at App Store Connect → Users and Access → Integrations → App Store Connect API, download the .p8, and add env vars to the MCP config:
{
"mcpServers": {
"ios-ship-doctor": {
"command": "ios-ship-doctor-mcp",
"env": {
"ASC_KEY_ID": "XXXXXXXXXX",
"ASC_ISSUER_ID": "xxxxxxxx-xxxx-xxxx-xxxx-xxxxxxxxxxxx",
"ASC_PRIVATE_KEY_PATH": "/absolute/path/to/AuthKey_XXXXXXXXXX.p8"
}
}
}
}
The .p8 never enters the repo — it stays on your machine and is read at runtime.
</details>
Manual configuration
If you'd rather not use npm link, point your client at the built file directly (this shape works for Claude Code, Gemini CLI, Cursor, Windsurf and most others):
{
"mcpServers": {
"ios-ship-doctor": {
"command": "node",
"args": ["/absolute/path/to/ios-ship-doctor/dist/index.js"]
}
}
}
projectPath in any tool can be your repo root (with an ios/ folder) or the ios/ directory itself.
Use it in CI (no assistant required)
The same binary runs as a one-shot command that exits non-zero on blocking issues:
npx -y ios-ship-doctor-mcp preflight /path/to/app # human-readable
npx -y ios-ship-doctor-mcp preflight /path/to/app --json # machine-readable
Copy examples/preflight-gate.yml into your app repo's .github/workflows/ to block PRs that would fail App Store review.
How it works
- Required-reason APIs — scans first-party source for Apple's documented API signatures, then diffs against your declared
PrivacyInfo.xcprivacy. - App Store Connect auth — an ES256 JWT signed with Node's built-in
crypto(nojsonwebtokendependency), in the JOSE (IEEE-P1363) format Apple requires. - Safe by default — everything is read-only except
generate_privacy_manifestwithwrite=true.
Roadmap
- [x] Auto-fix tools (patch
Info.plist, generate privacy manifest) - [x] More preflight checks: launch screen, version sanity, deployment target
- [x] SwiftPM dependency scanning
- [x] CLI + CI mode
- [ ] Entitlements sanity (push, app groups, associated domains)
- [ ] Privacy nutrition-label cross-check against SDK data collection
- [ ] Draft reviewer replies from a rejection
Contributing
Issues and PRs welcome. Adding a check usually means one entry in src/knowledge.ts plus a small function in src/scanner.ts.
License
MIT © menansali
Recommended Servers
playwright-mcp
A Model Context Protocol server that enables LLMs to interact with web pages through structured accessibility snapshots without requiring vision models or screenshots.
Audiense Insights MCP Server
Enables interaction with Audiense Insights accounts via the Model Context Protocol, facilitating the extraction and analysis of marketing insights and audience data including demographics, behavior, and influencer engagement.
Magic Component Platform (MCP)
An AI-powered tool that generates modern UI components from natural language descriptions, integrating with popular IDEs to streamline UI development workflow.
VeyraX MCP
Single MCP tool to connect all your favorite tools: Gmail, Calendar and 40 more.
graphlit-mcp-server
The Model Context Protocol (MCP) Server enables integration between MCP clients and the Graphlit service. Ingest anything from Slack to Gmail to podcast feeds, in addition to web crawling, into a Graphlit project - and then retrieve relevant contents from the MCP client.
Kagi MCP Server
An MCP server that integrates Kagi search capabilities with Claude AI, enabling Claude to perform real-time web searches when answering questions that require up-to-date information.
Neon Database
MCP server for interacting with Neon Management API and databases
Exa Search
A Model Context Protocol (MCP) server lets AI assistants like Claude use the Exa AI Search API for web searches. This setup allows AI models to get real-time web information in a safe and controlled way.
Qdrant Server
This repository is an example of how to create a MCP server for Qdrant, a vector search engine.
E2B
Using MCP to run code via e2b.