# kellyvv/PhoneClaw

PhoneClaw turns phones into local AI agent runtimes with on-device models, native mobile Skills, LiveLand, and optional Mac Gateway inference.

Repository: https://github.com/kellyvv/PhoneClaw
Canonical: https://ross.abutalabs.com/products/phoneclaw
Homepage: https://kellyvv.github.io/PhoneClaw/
Language: Swift
License: Apache-2.0
License Family: permissive
Topics: gemma, ios, litert, local-agent, mobile-agent, on-device-ai, swift, agent-framework, ai-agent, local-ai, ollama, mobile-ai, on-device-agent, mobile-agent-framework, mobile-ai-agent-framework, phone-ai-agent, phone-agent-harness, phone-agent-loop, phone-harness, phone-loop
Last push: 2026-08-06T08:01:19+00:00

## Health v2 (maintenance only)
Score: 77/100 (v2, computed 2026-09-03T02:20:16.233290+00:00)
- activity 96, release rhythm 90, longevity 10
- inputs: {"age_days": 151, "days_push": 27, "days_rel": 71, "gap_med": 2.5, "n_releases_24m": 9}
- flags: young
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 1220, forks 166 (observed 2026-08-28T04:04:01.947378+00:00)

## What it is
PhoneClaw is a mobile-native local AI agent framework that turns phones into on-device agent runtimes, running Gemma models via LiteRT and MiniCPM-V on iOS. It provides native mobile Skills (Calendar, Reminders, Contacts, HealthKit, clipboard, image understanding, voice), a LiveLand Dynamic Island interaction mode, and optional LAN-based Mac Gateway inference with Ollama.

## Use cases
- run AI agents fully offline on my iPhone
- ask my phone about health data with a local model
- create reminders and calendar events via natural language on device
- build phone-first local AI agent workflows
- use on-device Gemma models with native iOS skills
- offload agent inference to a Mac over LAN
- interact with a local AI agent from the Dynamic Island

## When to choose
- you want a privacy-focused, fully offline AI agent on iOS
- you need native access to phone capabilities like HealthKit, Calendar, and Contacts from an agent
- you want on-device small model inference (Gemma via LiteRT) on mobile hardware
- you want optional LAN edge inference without cloud dependency

## When to avoid
- you need an Android agent runtime (this iOS runtime is unrelated to same-named Android projects)
- you need large frontier-model quality that phones cannot run locally
- you need a cross-platform or server-side agent framework
- you need cloud-hosted LLM APIs with broad tool ecosystems

## Facets
- artifact type: framework
- maturity: active
- function: agent-framework, llm-inference, machine-learning, speech-recognition, image-processing
- domain: artificial-intelligence, large-language-models, mobile-development, apple-ecosystem, privacy
- platform: -
- tags: on-device-ai, local-first, litert, gemma, ollama, livel-land, edge-inference, native-skills, healthkit, ai-agents, ios, swift, mobile, macos

## Member repositories
- kellyvv/PhoneClaw (main) score 77

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:04:01.947378+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-30T06:16:01.001344+00:00, confidence not recorded.
  - readme: https://github.com/kellyvv/PhoneClaw (fetched 2026-08-28T04:04:01.947378+00:00, sha 8e6d80573b0c)
  - homepage: https://kellyvv.github.io/PhoneClaw/ (fetched 2026-08-29T12:24:17.282880+00:00, sha 8775277dcc07)
- Data as of 2026-08-30T08:39:29.467469+00:00.