#2323rd of 43 in Code review and testing
solopi
Mobile app test automation with Android record/replay and an agent execution harness
- Stars
- 6.4k
- License
- Apache-2.0
- Last commit
- Nov 2026
- Last release
- Nov 2026
- Language
- Java
6,109 on 21 Apr 20266,366 on 10 Nov 2026
From GitHub's daily star counts.
Checked on GitHub.
Overview
SoloPi records and replays UI flows on real Android devices, usually without adding an SDK to the app under test. It also includes an Electron-style Desktop workbench and a Skill/Harness layer that re-checks the current page before each agent action. Results are judged separately from action success as passed, failed or not_tested, with screenshots, logs and reports saved per run.
Who it is for: Mobile QA engineers and teams wiring AI agents into device testing
Strengths
- Target apps usually need no SoloPi SDK; works on real devices
- Separates action execution from pass/fail, with passed, failed and not_tested verdicts
- Saves screenshots, logs, action receipts and structured reports for each run
- Apache-2.0 with Android app, Desktop and CLI entry points
Weaknesses
- Desktop control, Skill/Harness are Beta; iOS (WDA) and HarmonyOS NEXT are experimental
- Android performance plugins and recording are not ported to iOS or NEXT
- Multi-device control was never open-sourced and is unavailable
- Built-in workflow templates need specific apps, accounts and starting pages
What it needs
- Needs adb, Node.js 22.14.0 (Desktop), npm, Python 3.9+, Android device with USB debugging
What changed
From the release notes. No flag means the note does not say so, not that the upgrade is safe.
Also in Code review and testing
See all 43| Rank | Project | Score |
|---|---|---|
| 1 | playwrightCross-browser automation and end-to-end test framework for Chromium, Firefox and WebKit | 94 out of 100 |
| 2 | nxMonorepo build system with task caching, affected-only runs and CI integration | 93 out of 100 |
| 3 | open-code-reviewGo CLI that reviews Git diffs with an LLM agent and line-level comments | 92 out of 100 |
| #4 | bendDependent-typed language where AI-written code must prove rules in LAWS.bend | 88 out of 100 |
| #5 | maestroYAML-based UI and end-to-end test framework for Android, iOS and web | 88 out of 100 |