Phone-harness puts iPhones under agent control

// 4h agoNEWS

Phone-harness puts iPhones under agent control

phone-harness lets AI agents control real iPhones through macOS iPhone Mirroring, using OCR for screen understanding and HID events for taps, typing, scrolling, and gestures. The open-source project recently passed 1,000 GitHub stars.

// ANALYSIS

phone-harness turns mobile automation from an API integration problem into a practical computer-use workflow, though its macOS permissions and iPhone Mirroring dependency limit its reach today.

–Agents can operate essentially any iOS app without jailbreak, Xcode, or WebDriverAgent
–OCR plus screenshot verification gives agents a workable substitute for a mobile accessibility tree
–The editable harness lets agents add reusable helpers during execution instead of waiting for framework support
–macOS Accessibility and Screen Recording permissions create setup friction and meaningful security considerations
–No multitouch, Face ID, camera flows, or DRM support means it is powerful for workflows but not yet a universal mobile automation layer

// TAGS

phone-harnessagentcomputer-useautomationdevtoolopen-source

DISCOVERED

4h ago

2026-08-12

PUBLISHED

2d ago

2026-08-10

RELEVANCE

8/ 10

AUTHOR

shawn_pana

// KEEP READING

More AI developer news from the feed

EXPLORE FULL FEED

UPDATE23m ago

Seedance 2.5 Demonstrates Longer, Editable AI Video

ByteDance’s Seedance 2.5 generates coherent video clips up to 30 seconds with synchronized audio, multimodal references, stronger prompt adherence, and region-level editing. The video demonstrates multi-shot workflows and autonomous film creation.

INFRA33m ago

Groq Joins NVIDIA Cloud Partner Network

Groq has joined NVIDIA’s Cloud Partner network, reinforcing its position as a serious provider of low-latency AI inference infrastructure. The partnership gives developers another path to production-grade inference beyond conventional GPU cloud deployments.

BENCHMARK37m ago

Grok 4.5 Closes In On Frontier

SpaceXAI’s Grok 4.5 is emerging as a near-frontier model for coding, agentic tasks, and knowledge work, with strong benchmark results and substantially lower API pricing than leading rivals. Independent evaluations suggest it is competitive, though not yet a clear overall leader.