Android Test Pilot
- Updated April 21, 2026
- panicgit/android-test-pilot
android-test-pilot is a Claude Code plugin marketplace that automates Android app testing via static analysis, log coverage checks, and three-tier device testing for developers replacing screenshot-heavy mobile-mcp loops
About
android-test-pilot is a Claude Code plugin marketplace from panicgit/android-test-pilot that automates Android app testing with a text-first strategy cheaper and faster than screenshot-per-step mobile-mcp loops. Step 0 static analysis produces navigation_map.mermaid, api_scenarios.json, and view_state_map.json under .claude/app-map. Step 1 checks ATP_SCREEN, ATP_RENDER, and ATP_API logcat tag coverage and augments missing instrumentation. Step 2 runs markdown or JSON scenarios through a three-tier fallback: Tier 1 dumpsys activity, dumpsys window, and logcat text; Tier 2 uiautomator accessibility trees; Tier 3 screenshots only as last resort. The MCP server exposes five tools—atp_run_step, atp_dumpsys, atp_logcat_start, atp_logcat_read, atp_logcat_stop—plus inherited mobile-mcp device tools. Requirements are Node.js 18+, ADB, and Claude Code. Install via /plugin marketplace add panicgit/android-test-pilot with slash commands /atp:analyze-app, /atp:check-logs, /atp:run-test, and /atp:app-map. Reach for android-test-pilot when Android QA needs deterministic text-based state checks before spending image tokens on screenshots.
- 3-step flow: static analyze → log coverage check → 3-tier device test execution
- Tier 1 dumpsys+logcat text; Tier 2 uiautomator; Tier 3 screenshot last resort
- 5 MCP tools: atp_run_step, atp_dumpsys, atp_logcat_start/read/stop
- 4 slash commands: /atp:analyze-app, /atp:check-logs, /atp:run-test, /atp:app-map
- 3 ATP log tags: ATP_SCREEN, ATP_RENDER, ATP_API for text-based state detection
Android Test Pilot by the numbers
- Data as of Jul 7, 2026 (Skillselion catalog sync)
/plugin marketplace add panicgit/android-test-pilotAdd your badge
Show developers this marketplace is listed on Skillselion. Paste this into your README.
| Last updated | April 21, 2026 |
|---|---|
| Repository | panicgit/android-test-pilot ↗ |
How do you automate Android tests without screenshot loops?
Run android-test-pilot /atp:analyze-app and /atp:run-test to execute Android scenarios via dumpsys, logcat, and uiautomator with screenshot fallback only when needed.
Who is it for?
Android developers using Claude Code who want cheaper, faster device test automation via dumpsys and logcat text instead of per-step screenshot LLM analysis.
Skip if: iOS-only teams, projects without ADB access, or workflows that do not use Claude Code marketplace plugins.
What you get
navigation_map.mermaid, api_scenarios.json, view_state_map.json, augmented ATP_* logcat tags, and tier-labeled scenario pass/fail results.
- navigation_map.mermaid
- api_scenarios.json
- scenario pass/fail results
By the numbers
- Marketplace bundles 1 plugin with 5 dedicated MCP tools plus mobile-mcp tools
- 3-tier test strategy: text dumpsys+logcat, uiautomator, then screenshot
- 3 ATP log tags: ATP_SCREEN, ATP_RENDER, ATP_API; requires Node.js ≥ 18
Recommended Marketplaces
How it compares
Use android-test-pilot for text-first Android scenario testing in Claude Code; use raw mobile-mcp when screenshot-driven exploration is acceptable.
FAQ
How is Android Test Pilot faster than mobile-mcp?
Android Test Pilot uses Tier 1 dumpsys and filtered ATP_* logcat text for state detection, avoiding per-step screenshot image tokens; screenshots are only used in Tier 3 as a last resort for rendering verification.
What MCP tools does Android Test Pilot provide?
Android Test Pilot exposes five MCP tools—atp_run_step, atp_dumpsys, atp_logcat_start, atp_logcat_read, and atp_logcat_stop—plus inherited mobile-mcp screenshot and UI interaction tools.
What log tags does Android Test Pilot expect?
Android Test Pilot expects ATP_SCREEN for screen transitions, ATP_RENDER for view state changes, and ATP_API for API response logging, checked and augmented during the /atp:check-logs step.