Metavr Cli

metavr CLI Reference

What metavr Does

metavr (Meta VR CLI) is a CLI for Meta Quest and Horizon OS development. It wraps ADB and Meta developer APIs into higher-level commands for device management, app lifecycle, documentation lookup, screenshots, audio control, test-device preparation, MCP integration, and Perfetto performance analysis.

Keep this SKILL.md as the routing layer. For exact flags and complete command syntax, use the generated full CLI reference in docs/hzdb.md or run:

npx -y metavr --markdown-help

Installation

Requires Node.js 20 or newer.

Invoke via npx:

npx -y metavr --help
npx -y metavr --version

npx is the recommended path: it always pulls the latest published version, so you don't end up running an outdated global install. Examples in the rest of the documentation use the bare metavr command for brevity — substitute npx -y metavr for any of them.

Alternatively, install metavr globally using:

npm install -g metavr

If you install globally, update regularly with npm update -g metavr. Use npx -y metavr when you want the latest published version without depending on a global install.

Device Connection

Before using on-device commands:

1. Enable Developer Mode on your Quest (Settings > System > Developer)
2. Enable USB Debugging when prompted
3. Connect your Quest to your computer via USB-C

Verify the connection:

metavr device list

If no devices appear: try a different USB cable (data-capable, not charge-only), accept the USB debugging prompt on the headset, or connect wirelessly with metavr device connect .

Command Discovery

Use progressive disclosure:

1. Start with this skill for command groups and common workflows.
2. Open a focused reference file when the user is working in that area.
3. Use docs/hzdb.md or metavr --markdown-help for exact flags, arguments, and

less common subcommands.

Current top-level command groups:

| Group | Use for | |---|---| | adb / shell | Low-level ADB-compatible commands and direct device shell access | | app | Install, list, launch, stop, clear, inspect, and detect the foreground app | | asset | Search Meta's 3D asset library | | audio | Read volume, set volume, mute, and unmute device audio | | capture | Capture screenshots from a connected headset | | config | Read, write, reset, and list metavr configuration | | device | List/connect devices, wait for ADB state, inspect controllers, configure test devices, run health checks, and manage proximity | | docs | Search/fetch Meta Quest docs and API reference entries | | files | List, pull, push, remove, and create directories on the device | | log | View recent device logs; use adb logcat for advanced filters and streaming | | mcp | Start or install the metavr MCP server for AI tools | | perf | Capture, open, analyze, query, compare, and manage Perfetto traces |

MCP Server Mode

metavr includes a built-in MCP (Model Context Protocol) server with ~40 tools that enable AI agents to interact with Quest devices programmatically.

Add this command to any MCP configuration to use the MCP server:

metavr mcp server

The server exposes tools for device management, app lifecycle, performance trace analysis, documentation search, file operations, and 3D asset search.

Install MCP configuration into other AI tools:

metavr mcp install cursor
metavr mcp install claude-desktop
metavr mcp install vscode

For project-local agent setup, install the MCP config into the current repository:

cd your-project
metavr mcp install project

This is the best default when a coding agent already has repository access on the host machine and should be able to call metavr tools from that project.

Common Agent Workflow

If you are building a Quest-native developer tool or pairing a Quest headset UI with a host-side coding agent, keep the architecture simple:

• Let the host-side coding agent own repository access, file edits, builds, tests,

and metavr tool calls

• Let the headset app or browser experience act as a thin client for preview,

prompt capture, status, and approval

• Prefer project-local MCP installation with metavr mcp install project so the

integration is explicit and travels with the repository

# 1. Verify the current docs before coding against an API or workflow
metavr docs search "spatial sdk panel"
metavr docs fetch https://developers.meta.com/horizon/documentation/...

# 2. Build and deploy your app
./gradlew assembleDebug
metavr app install app/build/outputs/apk/debug/app-debug.apk
metavr app launch com.example.app

# 3. Observe on-device behavior
metavr log --tag MyApp
metavr capture screenshot -o latest.png

# 4. Prepare a stable test device when running repeatable tests
metavr device health-check
metavr device configure-testing setup
# ...run tests...
metavr device configure-testing restore

Relationship to ADB

metavr wraps ADB and provides higher-level commands. You do not need to use adb directly for most Quest development tasks. metavr handles device selection, provides structured output, and adds Quest-specific functionality (screenshots via metacam, Perfetto trace analysis, doc search) that raw ADB does not support.

If you need raw shell access to the device, use metavr shell or metavr adb shell.

Full Command Reference

Use docs/hzdb.md for generated command details. Regenerate it with:

npx -y metavr --markdown-help > docs/hzdb.md

References

For detailed usage guides with workflows, examples, and troubleshooting:

• [Device Management](references/metavr-device-management.md) — device commands, audio, screenshots, logs, test setup, health checks, shell access
• [App Management](references/metavr-app-management.md) — app lifecycle, foreground app detection, crash debugging, common log tags
• [Agent Workflows](references/metavr-agent-workflows.md) — project-local MCP install, docs verification, safety, thin-client architecture
• [Performance Tools](references/metavr-perf-tools.md) — trace capture, guided analysis, Perfetto UI, SQL queries, comparisons, GPU counters
• [Documentation Search](references/metavr-docs-search.md) — doc search, API search, category filtering

Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

• Author: meta-quest
• Source: meta-quest/agentic-tools
• License: Apache-2.0
• Homepage: https://developers.meta.com/horizon

Install and usage instructions live in the source repository linked above.