手机爪AI代理

Requirements

• macOS + Xcode 16 or later
• iOS 17.0 or later
• CocoaPods
• A real device with a developer account (Apple ID)

Model recommendation:

2. Install dependencies

pod install

Default Install Flow and Model Bundling

Built-in Skill Examples

Calendar: Create calendar events, query schedules, and analyze busyness or free time using natural language.

"Schedule a meeting at Hightech Park tomorrow at 2pm"

"What is on my calendar today?"

"How busy am I this week?"

Reminders: Set time-based reminders that fire a system push notification exactly on schedule.

"Remind me tonight at 8 to send the file to my boss"

Contacts: Search, save, update, or delete contacts with name, phone, company, email, and notes. Automatically deduped by phone number.

"Save Wang's number 13812345678, he's from Bytedance"

"Check Sarah Lee's phone number"

Clipboard: Read and write the system clipboard. Useful as a data relay in multi-step tasks.

"Copy that text to the clipboard"

Translate: Translate between any pair of languages, with automatic source detection.

"Translate that last line into Japanese"

Health Data: Read HealthKit steps, distance, calories, heart rate, sleep, and workout records after user authorization. All data stays on-device.

"How many steps did I take today?"

"How did I sleep last night?"

"How are my steps this week?"

"What's my resting heart rate?"

Web Search: When explicitly requested, search public webpages or read a URL, then summarize realtime information into an answer.

"Search the web for today's AI news"

"Read and summarize this page: https://example.com"

Quick Start

Recommended install: TestFlight. After installing, download a model in Model Settings, then enable the Skills you need.

Building from source requires macOS + Xcode 16, iOS 17+, CocoaPods, a real device, and an Apple ID.

Demo

3. Optional: pre-download a model locally

The default recommended flow is now:

1. Install the app shell to the iPhone from Xcode
2. Open the app
3. Go to Model Settings
4. Download Gemma 4 E2B or Gemma 4 E4B directly on the phone

You only need the Models/ directory on your Mac if you want to bundle a model inside the app itself.

Gemma 4 now runs on LiteRT-LM: each model is a single .litertlm file in place of an MLX weight directory. Install the Hugging Face CLI first:

```bash brew install hf

5. Configure signing and run

1. In Xcode, select the PhoneClaw target
2. Open Signing & Capabilities
3. Set your Team
4. Change the Bundle Identifier to a unique value
5. Connect your iPhone and press ⌘R

On first install, if prompted to trust the developer certificate: Settings → General → VPN & Device Management → Trust

Option A — Shell app + on-device model download

This is now the default recommended setup.

Advantages:

1. Much smaller install size from Xcode
2. Faster first-time app installation from the Mac
3. Users can choose E2B or E4B directly on the phone

By default, the project builds a lightweight app while model files download on device or get copied explicitly during custom builds.

Option B — E2B only

1. Keep Models/gemma-4-E2B-it.litertlm, remove Models/gemma-4-E4B-it.litertlm
2. In Xcode's Project Navigator, delete the unused model file reference and choose Remove Reference
3. In PhoneClaw > Build Phases > Copy Bundle Resources, make sure gemma-4-E2B-it.litertlm is included as a single file resource
4. Edit allModels in LLM/Models/PredefinedModels.swift to only include the models actually shipped (otherwise the settings page will show options that don't exist)

Option C — Both E2B and E4B

Download both models:

brew install hf
mkdir -p ./Models
hf download litert-community/gemma-4-E2B-it-litert-lm gemma-4-E2B-it.litertlm --local-dir ./Models
hf download litert-community/gemma-4-E4B-it-litert-lm gemma-4-E4B-it.litertlm --local-dir ./Models

Then add both .litertlm files into Xcode's Copy Bundle Resources.

1. More iOS native APIs

• [ ] File and directory access
• [x] Image picking, description, and Q&A
• [ ] Photo library reading, organization, and search
• [ ] Notes
• [x] Reminder due-time alerts
• [ ] General local notifications
• [ ] Maps and location
• [x] URL webpage reading and context passing
• [ ] Safari / URL Scheme handoff to external apps
• [x] Contacts search, create, update, and delete
• [x] Calendar creation, schedule reading, busyness, and free-time analysis
• [x] Reminder creation
• [x] Read-only HealthKit analysis

References

• Hugging Face CLI documentation
• Hugging Face download guide
• Gemma 4 E2B LiteRT model
• Gemma 4 E4B LiteRT model
• Gemma 4 E2B (ModelScope mirror)
• Gemma 4 E4B (ModelScope mirror)
• MiniCPM-V 4.6 model
• OpenBMB MiniCPM-V iOS Demo

FAQ

How do I trigger permission dialogs after install? Permission dialogs appear when a Skill reaches the related system API call. After a previous denial, re-enable the permission in system Settings.

Why does the model fail to load after switching? Verify that the model file name matches allModels in LLM/Models/PredefinedModels.swift, that the model has finished downloading on-device if you are using the shell-only install flow, or that it was actually included in the app bundle if you are shipping it built-in, and that the device has enough memory.

Why does creating a reminder fail? The latest code first attempts to reuse an existing writable reminder list. If none is found, it tries to automatically create a PhoneClaw list. If that also fails, the system reminder source itself is likely read-only.

How do I make the iPhone discover the Mac client? Make sure PhoneClawGateway.app is running and macOS has allowed Local Network permission. The iPhone and Mac must be on the same LAN. If discovery still fails, check that the macOS firewall allows the app to accept incoming connections, or rebuild the app with bash MacGateway/build-app.sh and open the generated app again.