lazy-mobile-mcp
Local MCP server for Android and iOS mobile automation and performance telemetry, enabling device control (screenshot, tap, swipe, input, app launch) and metric collection (CPU, memory, launch time) via ADB, simctl, and WebDriverAgent with SQLite session history.
README
lazy-mobile-mcp
Local Model Context Protocol (MCP) server for Android and iOS mobile automation with performance telemetry.
Control real devices and simulators using screenshot, tap, swipe, input_text, launch_app, and collect baseline metrics (cpu, memory, launch time) with session history in SQLite.
Keywords: MCP server, mobile automation, Android ADB automation, iOS simulator automation, WebDriverAgent, app performance telemetry.
Why This Project
- Build a local MCP bridge for AI clients over
stdio. - Automate Android via ADB and iOS via
simctl/devicectl/ WDA. - Keep operations traceable with
trace_idin responses and logs. - Persist sessions, samples, artifacts, and audit events in SQLite for reproducibility.
Features
stdioMCP transport for local AI tooling.- Single active device model (
select_device) with optional per-calldevice_id. - Android adapter via ADB CLI.
- iOS adapter via Xcode tools with macOS guard and graceful degradation.
- WDA endpoint auto-discovery for iOS interactive actions.
- Risk-tiered approval policy for interactive actions.
- JSON logging and unified error contract.
- SQLite persistence for
sessions,perf_samples,artifacts,audit_logs.
Tool Index
mobile.list_devicesmobile.select_devicemobile.get_capabilitiesmobile.screenshotmobile.tapmobile.swipemobile.input_textmobile.launch_appmobile.stop_appmobile.start_perf_sessionmobile.stop_perf_sessionmobile.get_perf_samples
Architecture
- TypeScript MCP server + worker (tool contracts, validation, policy, trace ID)
- Android adapter (
adb) - iOS adapter (
simctl,devicectl, WDA) - SQLite storage (
artifacts/mobile.db)
Prerequisites
- Node.js 20+
- Android:
adbinPATH - iOS (optional): macOS +
xcrun(simctl/devicectl) - For iOS interactive actions (
tap/swipe/input): reachable WebDriverAgent endpoint
Install
npm install
Install From npm
npm install lazy_mobile_mcp
Quick Start (Codex)
Requires codex CLI in PATH.
Published package:
npx -y lazy_mobile_mcp@latest setup-codex
Verify registration:
codex mcp get lazy-mobile-mcp
Then open a new Codex session and call mobile.list_devices.
Current local checkout:
node bin/lazy-mobile-mcp.js setup-codex --local --name lazy-mobile-mcp-local
Codex One-Command Setup (Advanced)
Optional overrides:
npx -y lazy_mobile_mcp@latest setup-codex \
--name lazy-mobile-mcp \
--sqlite-path "$HOME/.codex/mcp-data/lazy-mobile/mobile.db" \
--adb-bin adb \
--wda-base-url http://127.0.0.1:8100 \
--device-allowlist emulator-5554 \
--approval-policy high
Local checkout with the same overrides:
node bin/lazy-mobile-mcp.js setup-codex \
--local \
--name lazy-mobile-mcp-local \
--sqlite-path "$HOME/.codex/mcp-data/lazy-mobile/mobile.db" \
--adb-bin adb
Run with npx:
npx -y lazy_mobile_mcp@latest
Global install:
npm install -g lazy_mobile_mcp
lazy-mobile-mcp
Run
Development:
npm run dev
Production:
npm run build
npm start
Configuration
SQLITE_PATH(defaultartifacts/mobile.db)DEVICE_ALLOWLIST(comma-separated)MOBILE_APPROVAL_POLICY(off|high|medium, defaultoff)LOG_LEVEL(debug|info|warn|error)ADB_BIN(defaultadb)WDA_BASE_URL(optional override for iOS WDA endpoint)
If WDA_BASE_URL is not set, the adapter probes common local endpoints (127.0.0.1 / localhost, ports 8100/8101/8200/8201 + local listening ports).
Security Model
lazy-mobile-mcp is a local-first MCP server. It runs over stdio with the permissions of the local OS user that starts it; it is not a hardened OS or container sandbox.
Current runtime controls:
DEVICE_ALLOWLISTcan restrict operations to known device IDs.- All tool inputs are schema-validated before adapter execution.
- Tool attempts are recorded in SQLite
audit_logswithtrace_id, tool name, risk level, device ID, and result code. MOBILE_APPROVAL_POLICY=highrequiresconfirm=trueand a non-emptyreasonfortap,swipe,input_text,launch_app, andstop_app.MOBILE_APPROVAL_POLICY=mediumalso requires confirmation forscreenshotandstart_perf_session.
The confirm / reason fields are caller confirmation fields. They do not prove a human approved the action unless the MCP host or caller enforces a human-in-the-loop approval flow.
iOS Capability Notes
- Simulator: screenshot + launch/stop + WDA interactive actions.
- Physical device: launch/stop via
devicectl; screenshot and interactive actions via WDA. - Non-macOS host: iOS tools return
ERR_IOS_UNAVAILABLE_ON_HOST.
Testing
npm test
Recommended local MCP smoke:
node dist/cli.js --help
node dist/cli.js setup-codex --help
Recommended Servers
playwright-mcp
A Model Context Protocol server that enables LLMs to interact with web pages through structured accessibility snapshots without requiring vision models or screenshots.
Magic Component Platform (MCP)
An AI-powered tool that generates modern UI components from natural language descriptions, integrating with popular IDEs to streamline UI development workflow.
Audiense Insights MCP Server
Enables interaction with Audiense Insights accounts via the Model Context Protocol, facilitating the extraction and analysis of marketing insights and audience data including demographics, behavior, and influencer engagement.
VeyraX MCP
Single MCP tool to connect all your favorite tools: Gmail, Calendar and 40 more.
graphlit-mcp-server
The Model Context Protocol (MCP) Server enables integration between MCP clients and the Graphlit service. Ingest anything from Slack to Gmail to podcast feeds, in addition to web crawling, into a Graphlit project - and then retrieve relevant contents from the MCP client.
Kagi MCP Server
An MCP server that integrates Kagi search capabilities with Claude AI, enabling Claude to perform real-time web searches when answering questions that require up-to-date information.
E2B
Using MCP to run code via e2b.
Neon Database
MCP server for interacting with Neon Management API and databases
Qdrant Server
This repository is an example of how to create a MCP server for Qdrant, a vector search engine.
Exa Search
A Model Context Protocol (MCP) server lets AI assistants like Claude use the Exa AI Search API for web searches. This setup allows AI models to get real-time web information in a safe and controlled way.