ARTEMIS is an open-source framework from Google that turns natural-language instructions into reliable Android automation. Released under the Apache-2.0 license, it has attracted more than 13,300 stars on GitHub and was actively developed in 2026. Rather than brittle recorded scripts, ARTEMIS drives end-to-end workflows from plain-English descriptions, and reports a success rate above 99% on the AndroidWorld benchmark.
Mobile UI testing is notoriously fragile: selectors break, flows change, and maintaining scripted tests eats engineering time. ARTEMIS approaches the problem with AI agents that look at the screen, reason about the instruction, and act step by step until the workflow completes. It captures logs and screenshots along the way, so failures come with evidence, and it is designed to plug straight into the AI coding assistants teams already use โ including Antigravity, Codex and Claude Code โ letting those tools not just write Android code but actually verify it on a device or emulator.
Because the instructions are natural language, the same description keeps working as an app’s layout evolves, removing the constant maintenance treadmill of locator-based test suites. Teams can treat a written scenario as the source of truth and let ARTEMIS handle the taps, swipes, text entry and assertions on real app flows, from onboarding and login to checkout-style multi-screen journeys.
ARTEMIS is free and open source. You run it yourself against your own devices or emulators; any model costs depend on the AI assistant or API you connect to it.
ARTEMIS is built for Android developers, QA engineers and agentic-coding adopters who want natural-language end-to-end testing and device automation that survives UI changes better than traditional script-based frameworks.