AIAny
Icon for item

ARTEMIS

Turns natural-language instructions into end-to-end Android automation that drives real devices and captures Logcat and screenshots. Uses a multimodal “Dynamic‑First, Coordinate‑Fallback” locator and a native MCP server for IDE integration, supporting fast Flash (3–5s/step) and verifiable Pro workflows.

Introduction

Mobile UI automation breaks quickly because selectors, pixel layouts and transient system UI change frequently. ARTEMIS takes a different stance: instead of brittle selectors it combines accessibility hierarchies, OCR and visual models with a "Dynamic‑First, Coordinate‑Fallback" locator pattern and exposes a native Model Context Protocol (MCP) server so AI IDEs can treat real phones like first‑class test targets. That design aims to make long‑horizon, cross‑app workflows (and human‑style exploratory testing) reliable and diagnosable.

What Sets It Apart
  • Multimodal grounding with safety checks — accessibility trees + OCR + visual models are used before falling back to coordinates, so actions remain resilient across Compose/Flutter/canvas UIs and layout drift. This reduces maintenance compared with XPath/ID selector approaches.
  • Native MCP integration for AI IDEs — a built‑in MCP server and a rules.md testing mindset let agents (Codex, Claude Code, Antigravity, Windsurf, etc.) drive devices, collect Logcat, screenshots and structured traces without handcrafted adapter glue. So AI assistants can plan, execute and fetch diagnostics end‑to‑end.
  • Two execution profiles for different needs — Flash: a reactive single‑model loop (~3–5s per step) for fast deterministic tasks; Pro: a multi‑agent planner/operator workflow (15–40s per step) with pre‑execution checks, checkpoints and final verification for longer, safety‑sensitive scenarios.
  • Benchmarked robustness — demonstrated >99% completion on the AndroidWorld multi‑step benchmark, indicating strong cross‑app reliability for scripted multi‑step tasks.
Who It's For and Tradeoffs

Great fit if you need reproducible, cross‑app Android test automation driven by natural language or AI agents, want integrated diagnostics (Logcat/screenshots/traces), or plan long‑running exploratory/stability tests (Pro supports 10+ hour runs). It also suits teams building AI‑driven developer workflows inside IDEs that can mount MCP servers.

Look elsewhere if you only target iOS (iOS support is on the roadmap), need a zero‑footprint remote testing service (ARTEMIS installs an optional on‑device accessibility helper and expects USB debugging/emulator access), or require purely server‑side automation without any device connection. It’s optimized around device‑connected, privacy‑aware automation rather than cloud‑hosted device farms.

If you evaluate it, expect some initial device setup (USB debugging or emulator, optional helper install) and choose Flash for throughput and Pro for verifiability and long‑horizon workflows.

More Items

GitHub
AI Video2026

Turns Claude Code or Codex into an agent skill that storyboards, animates, and sound-designs cinematic product promos from your real screenshots using Remotion. Packs tuned shot recipe cards, 150+ motion previews, a production-ready template, Remotion components, and audio assets for web/desktop product films.

GitHub
AI Agent2026

Helps produce and revise natural, readable Japanese for work documents by applying a 12-item writing constitution and sudachipy-powered lint to detect and remove “AI-smelling” patterns; includes diagnostic scoring and iterative rewrite loops.

GitHub

Self-hosted web and phone app to plan weekly routines, run guided workouts, and log every set and bodyweight with passkey sign-in and local JSON storage. Syncs across devices, runs via Docker or a standalone Android APK, optional AI coach and local MCP bridge, no telemetry.