Everyone is building the same thing…
Lately, it looks like everybody is trying to build more or less the same thing. AI-enabled mobile testing and building tools. Too many to try and these are just the ones that I’m aware of:
- Tester Army - announced just today
- agent-device - pretty fun project from @thymikee that seems to be basis for their SaaS
- Qampanion - seeing their ads everywhere
- Autosana
- MobAI - seems like webview based desktop app
- Maestro Cloud - I used it before both locally and in the cloud, they seem to be reinventing few things towards the agentic approach
- Argent - saw this on Twitter a few times
- SimDeck from NativeScript
- Patrol MCP Release from LeanCode
- Quash
- Mobile MCP - played with that a bit
- QA.tech
I ported agent-device to Dart as agent-device-dart, but still not happy with the performance. The intention is to have a CLI tool that agent can use to run the app, but in the end it can also generate standard Dart tests (example) to be run as part of the CI.
The dream ✨ is to have mostly automated mobile development flow. Either have a QA tool that AI agent can operate to validate in-app flows, or a dev tool that allows AI agent to iterate on a mobile app feature end-to-end.
One big problem I’m seeing is the flakiness of the underlying “plumbing”.
Accessibility tooling in simulators and devices is not reflective of the actual VoiceOver capabilities. Not every device feature is enabled on Android emulator or iOS simulator. Emulators and debug builds are slow.
The other way is to hook up to a given framework’s utilities, like Flutter’s VM service (e.g. what patrol is doing) or React Native’s dev tools (metro), but that is not cross-platform and also not really reflective of the real user experience e.g. in release mode.
I’m watching this space and would love to see something that can work well across emulators and real devices, that is accessible to both QA and devs, and that is able to deal with the flakiness of mobile E2E testing.
Have you solved it? If so, please share it.