google/artemis

ARTEMIS turns natural-language instructions into reliable Android automation. It automates end-to-end workflows, captures logs, and integrates seamlessly with AI coding assistants such as Antigravity, Codex, and Claude Code. It also achieves 99%+ success rate on AndroidWorld Benchmark.

What it solves

Artemis allows AI assistants and automated test suites to interact with real Android devices as a human would. It solves the difficulty of automating complex, multi-step mobile workflows—especially those involving custom interfaces (like Flutter or Compose) that traditional automation tools struggle to target—by using multimodal AI to perceive and navigate the UI via natural language instructions.

How it works

Artemis uses a reactive "observe-and-act" loop powered by multimodal models (such as Gemini, Claude, or GPT-4o). It identifies UI elements by combining accessibility hierarchies, OCR, and visual coordinates. To ensure reliability, it employs a "Dynamic-First, Coordinate-Fallback" locator pattern.

The system offers two execution modes:

  • Flash Profile: A fast, token-efficient loop (3–5s per step) for routine tasks, using history compression to maintain context.
  • Pro Profile: A sophisticated multi-agent graph involving a Planner (to create milestones), an Operator (to execute actions), and a Checker (to verify goals). It includes a "Safety Net" that checks targets before every action to prevent errors.

It integrates with IDEs via the Model Context Protocol (MCP), allowing AI coding assistants to drive physical devices, collect Logcat output, and take screenshots directly from the editor.

Who it’s for

  • Mobile QA Engineers: For creating autonomous stability and exploratory tests.
  • Android Developers: For reproducing bugs and automating repetitive device setup tasks.
  • AI Agent Developers: Who need a reliable way to let LLMs interact with mobile operating systems.

Highlights

  • SOTA Performance: Achieved 99%+ task completion on the AndroidWorld benchmark.
  • MCP Native: Direct integration with AI IDEs like Claude Code, Windsurf, and Antigravity.
  • Cross-App Automation: Capable of executing workflows across multiple different Android applications.
  • Multimodal Targeting: Supports element indices, coordinates, and visual locating for custom UI frameworks.
  • Flexible Interface: Accessible via a Web Visual Console, Developer CLI, Python SDK, or MCP server.

Related

  • Project
    openclaw/Peekaboo

    Peekaboo is a macOS‑only CLI and menu‑bar app that captures the screen, reads the accessibility tree, and performs native UI actions (click, type, press, etc.). It includes an LLM‑driven “agent” that can translate natural‑language instructions into sequences of these commands, and it can be exposed to external AI clients via a Model‑Control‑Protocol (MCP). Install via Homebrew, npm, or a signed DMG; configure model providers in `~/.peekaboo`. Ideal for GUI automation, testing, and building AI agents that control the desktop.

  • Project
    mavlink/qgroundcontrol

    A cross-platform Ground Control Station for MAVLink-enabled drones that provides full flight control, mission planning, and vehicle configuration.

  • Project
    RLinf/RLinf

    RLinf is an open‑source, PyTorch‑compatible reinforcement‑learning infrastructure for large‑scale embodied and agentic AI. It ties together many simulators and real‑world robot platforms, supports a wide range of accelerators, and provides ready‑made pipelines for fine‑tuning huge vision‑language, diffusion, and MoE models with RL algorithms such as PPO, GRPO, SAC, and newer flow‑transformation methods. The library includes extensive system optimizations (e.g., FUSCO) and dozens of example recipes, making it suitable for researchers and engineers who need a unified, scalable backbone for training and deploying RL agents in both simulation and the real world.

  • Project
    PetoiCamp/OpenCat-Quadruped-Robot

    An open-source framework for quadruped robots that handles gait coordination and servo control to enable easier development of AI-powered robotic pets.