What is PhoneClaw?
PhoneClaw is a mobile-native local AI Agent framework for phones and edge devices. Its iOS runtime runs on-device models and native iOS Skills on the phone.
Core facts
PhoneClaw in concrete terms.
PhoneClaw is easiest to understand as a phone runtime: local models, native Skills, permission-scoped data access, and optional LAN inference.
What can it do?
Use natural language for Calendar, Reminders, Contacts, Clipboard, HealthKit summaries, image understanding, voice and LIVE mode, LiveLand, translation, Web Search, and Mac Gateway.
How does data flow?
Chat, images, and personal data stay on the device by default. Web Search, URL reading, and Mac remote inference run when the user selects those capabilities.
Why mobile-native?
The runtime handles mobile memory budgets, resumable model downloads, scoped permissions, Skill routing, phone entry points, Live interaction, and LAN-based edge inference.
Phone runtime
Built around mobile runtime constraints.
PhoneClaw brings local inference, native data sources, permission prompts, model state, and Live interaction into one mobile Agent runtime.
Phone as runtime
The phone stays the center of the agent experience: chat, voice, camera, widgets, Shortcuts, Control Center, Dynamic Island, and LiveLand all route back to the same local task flow.
Native capability layer
Skills connect natural language to Calendar, Reminders, Contacts, Clipboard, HealthKit, translation, images, and explicit Web Search through mobile OS permissions.
Edge extension
A paired Mac can provide local-network inference for heavier tasks while the phone keeps the interaction surface, Skills, permissions, and task state.
Product
Designed around real phone work.
PhoneClaw treats the phone as the agent runtime for local inference, native Skills, permission checks, model downloads, memory budgeting, and Live interaction modes.
Native iOS Skills
Calendar, Reminders, Contacts, Clipboard, HealthKit, translation, images, and explicit Web Search.
Live and LiveLand
Voice, camera-aware LIVE mode, widgets, Shortcuts, Control Center, and Dynamic Island status.
Mobile runtime work
Model switching, resumable downloads, cache cleanup, history trimming, and local memory policy.
Mac as an edge node
Pair a Mac over LAN to use Ollama, Codex CLI, or Antigravity CLI while the phone keeps the agent experience.
Agent stack
A phone-first runtime from input to execution.
The framework combines mobile entry points, Skill-aware planning, native iOS tools, local models, and optional LAN inference into one runtime.
Mobile commands
Text, voice, images, LIVE camera, widgets, Shortcuts, Control Center, and LiveLand entry points.
Skill-aware planning
File-driven Skills, scoped tool lists, argument extraction, clarification, and multi-turn task handling.
Native iOS tools
Calendar, Reminders, Contacts, Clipboard, HealthKit, Web Search, and device-local summaries behind permission gates.
On-device or LAN
LiteRT local models for the fully offline path, with optional Mac Gateway for heavier models on a trusted local machine. PhoneClaw's on-device runtime is Google LiteRT (running Gemma 4 E2B / E4B) and MiniCPM-V 4.6 — not Apple MLX.
Data flow
Privacy claims mapped to product behavior.
The privacy page expands this into a full table. The summary below keeps the homepage specific and crawlable.
Stays on device
Default inference, native Skill calls, Calendar, Reminders, Contacts, Clipboard, HealthKit summaries, images, and chat.
Explicit web actions
Realtime Web Search and opening a public URL send the request to the search provider or target website.
Optional Mac inference
When paired, the request goes to your Mac on the LAN. Ollama stays there; CLI providers follow their own policies.
Mac Gateway
Keep the phone native. Move heavy inference to your Mac.
PhoneClaw Gateway turns a Mac on the same Wi-Fi into an optional local inference source. The phone keeps the chat UI, Skills, permission model, and task flow.
Deep dives
Technical proof for the main claims.
These pages give developers concrete implementation detail behind the product claims.
Mobile-native Agent framework
How PhoneClaw is designed from mobile permissions, memory budgets, phone entry points, and edge inference.
Gemma 4 on phone
LiteRT-LM, MiniCPM-V 4.6, mobile memory limits, and how model choice shapes local agent workflows.
Native iOS Skills
File-driven Skills, tool allowlists, permission boundaries, and local tool calling on the phone.
LiveLand and LIVE mode
Dynamic Island, widgets, voice, camera, and mobile task status in PhoneClaw's mobile runtime.
PhoneClaw FAQ
Concise answers about product identity, privacy, mobile agent architecture, models, Skills, and Mac Gateway.
Languages
Localized entry pages for Chinese and Japanese readers.
PhoneClaw ships product localization in zh/en/ja, with dedicated landing pages for readers who prefer Chinese or Japanese.
中文入口
中文介绍 PhoneClaw 的手机端 Agent 运行时、端侧模型、原生 Skills、隐私边界和 Mac Gateway。
日本語ページ
Japanese product overview for the local phone runtime, on-device models, native Skills, data flow, and Mac Gateway.
Project context file
A compact machine-readable summary of the project, models, Skills, data flow, and public links.