claude-in-mobile
claude-in-mobile is an MCP server (and also a standalone CLI) that lets an AI client control devices and apps for automation: Android via ADB, iOS Simulator via simctl and optional WebDriverAgent for UI inspection/element tapping, Desktop via a Compose Multiplatform companion, and Aurora OS via audb (with additional on-device Python for gestures). It exposes “tools” such as device listing, screenshots (including annotated screenshots), UI interactions (tap/swipe/input), app control, permissions management, and log/system information.
Score Breakdown
⚙ Agent Friendliness
🔒 Security
No auth model is described for the MCP server; it appears to rely on local execution. Because it can control devices, grant/revoke permissions, run shell commands (Android), and read logs, it has high potential impact if run in an unsafe context. Dependency hygiene and secret-handling details are not provided; assume default risk for unpublished implementation details. TLS is not applicable because the interface is stdio/local rather than network-based, but if any HTTP connections are made (e.g., WDA port 8100+), transport security guidance is not described.
⚡ Reliability
Best When
You want an agent-driven, cross-platform automation layer that speaks MCP/stdio to your AI client and you can provide the required local tooling (ADB/Xcode/Appium/WDA/audb/companion build).
Avoid When
You need enterprise auth, multi-tenant isolation, or strict compliance guarantees out of the box, or you cannot tolerate the operational complexity of device/simulator dependencies.
Use Cases
- • End-to-end UI test automation driven by natural language
- • Debugging mobile apps by inspecting UI and device logs
- • Automating repetitive mobile-device workflows (tap/type/navigation)
- • Cross-platform smoke testing across Android + iOS Simulator + desktop apps
- • Granting/revoking mobile app permissions during test setup
- • Screenshot capture for model-based UI understanding (including annotated screenshots)
Not For
- • Production-grade remote device control over untrusted networks without strong operational security controls
- • Controlling physical iOS devices (explicitly iOS Simulator only, per README)
- • Use where you need a stable, documented public HTTP/REST API contract (it is primarily MCP/stdio-based)
- • Workflows requiring guaranteed idempotency and transactional semantics (device actions are inherently stateful)
Interface
Authentication
The README does not describe user authentication/authorization for the server itself. Access is effectively local-process based (run via npx/CLI and MCP stdio), so security depends on who can execute the server and the underlying device tooling access.
Pricing
No pricing information in provided content; appears to be open-source (MIT) and self-hosted/local.
Agent Metadata
Known Gotchas
- ⚠ Statefulness: UI actions (tap/type/swipe/launch) are not idempotent; agents may need to confirm current screen before acting.
- ⚠ iOS element interactions depend on WebDriverAgent; without WDA, some element-based tools may fail or be limited.
- ⚠ Platform capability differences: some tools are explicitly not supported on certain targets (e.g., input_text on Aurora, desktop app install/stop/launch limitations as listed).
- ⚠ Device/environment dependencies: ADB connectivity, simulator boot state, WDA port availability (8100+), audb/SSH and on-device Python for Aurora gestures.
- ⚠ screenshot size/latency: annotated/smart screenshots can add processing time; agents should handle timeouts.
Alternatives
Full Evaluation Report
Comprehensive deep-dive: security analysis, reliability audit, agent experience review, cost modeling, competitive positioning, and improvement roadmap for claude-in-mobile.
AI-powered analysis · PDF + markdown · Delivered within 30 minutes
Package Brief
Quick verdict, integration guide, cost projections, gotchas with workarounds, and alternatives comparison.
Delivered within 10 minutes
Score Monitoring
Get alerted when this package's AF, security, or reliability scores change significantly. Stay ahead of regressions.
Continuous monitoring
Scores are editorial opinions as of 2026-03-30.