Mirroir – AI iPhone Control via macOS iPhone Mirroring

TL;DR

TL;DR: Mirroir is an open-source MCP server (Apache-2.0) that lets any AI agent control a real iPhone through macOS iPhone Mirroring — giving it screenshot access, tap/swipe gestures, and app control with 33 tools.

Source and Accuracy Notes

⚠️ This section is MANDATORY. All links must be verified from actual source, not guessed.

• Project page: mirroir.dev ← visited and verified
• Source repository: github.com/jfarcand/mirroir-mcp ← read README end-to-end
• License: Apache-2.0 (verified via LICENSE file)
• npm package: npmjs.com/package/mirroir-mcp — v0.35.0
• HN launch thread: news.ycombinator.com/item?id=47049130
• Source last checked: 2026-07-15 (README commit main)

What Is Mirroir?

Mirroir is an MCP server that bridges any MCP-compatible AI agent to a real iPhone via macOS iPhone Mirroring. The agent sees the screen via screenshot/OCR, decides what to do, then executes taps, swipes, and keystrokes through the server. No scripts, no coordinates — just intent.

“Give your AI eyes, hands, and a real iPhone.”

The project is not a GitHub repo that wraps an existing tool. It is a standalone MCP server built on macOS iPhone Mirroring APIs, with 33 tools covering screenshots, touch input, system gestures, app launching, and recording.

Key facts

• Package: mirroir-mcp on npm (v0.35.0)
• License: Apache-2.0
• Stars: 131
• Requirements: macOS 15+, iPhone connected via iPhone Mirroring
• Client support: Claude Code, Cursor, GitHub Copilot, OpenAI Codex

Setup Workflow

Step 1: Install Mirroir

Choose one of three install methods:

Install script (recommended):

/bin/bash -c "$(curl -fsSL https://mirroir.dev/get-mirroir.sh)"

Or via npx:

npx -y mirroir-mcp install

Or via Homebrew:

brew tap jfarcand/tap && brew install mirroir-mcp

Step 2: Grant Permissions

The first time you take a screenshot, macOS will prompt for:

• Screen Recording permission
• Accessibility permission

Grant both in System Settings → Privacy & Security → Screen Recording and Accessibility.

Step 3: Connect Your iPhone

Enable iPhone Mirroring on your Mac (macOS 15+):

1. Open Control Center on your iPhone
2. Tap iPhone Mirroring and select your Mac
3. Confirm the connection on both devices

Step 4: Add to Your AI Client

Claude Code

claude mcp add --transport stdio mirroir -- npx -y mirroir-mcp

Cursor

Add to .cursor/mcp.json in your project root:

{
  "mcpServers": {
    "mirroir": {
      "command": "npx",
      "args": ["-y", "mirroir-mcp"]
    }
  }
}

GitHub Copilot (VS Code)

Search @mcp mirroir in the Extensions view, or add to .vscode/mcp.json:

{
  "servers": {
    "mirroir": {
      "type": "stdio",
      "command": "npx",
      "args": ["-y", "mirroir-mcp"]
    }
  }
}

OpenAI Codex

codex mcp add mirroir -- npx -y mirroir-mcp

Or add to ~/.codex/config.toml:

[mcp_servers.mirroir]
command = "npx"
args = ["-y", "mirroir-mcp"]

Step 5: Verify the Connection

Run this in your AI client’s terminal to confirm Mirroir is connected:

# Ask your agent to list available tools
# It should show mirroir tools: screenshot, describe_screen, tap, swipe, etc.

How It Works

Every interaction follows the same observe, reason, act loop:

1. describe_screen — captures the screen and returns every UI element with tap coordinates (the AI’s “eyes”)
2. LLM reasoning — decides what action to take next (the “brain”)
3. tap, type_text, swipe — execute the action (the “hands”)
4. Loop back to step 1

Beyond basic taps, the server exposes 33 tools including:

| Tool | What it does | |------|-------------| | screenshot | Capture the iPhone screen as base64 PNG | | describe_screen | Analyze screen, return UI elements with tap coordinates | | tap / double_tap / long_press | Tap at coordinates | | swipe | Swipe between two points | | drag | Slow sustained drag for sliders, icons | | type_text | Send keystrokes to the iPhone | | press_key | Send special keys (return, escape, tab, arrows) with modifiers | | launch_app | Open app by name via Spotlight | | open_url | Open URL in Safari | | press_home / press_app_switcher / spotlight | System gesture controls | | press_back | Navigate back by OCR-tapping the back chevron | | scroll_to | Scroll until target text becomes visible | | reset_app | Force-quit an app via App Switcher | | set_network | Toggle connectivity through Settings | | measure | Time how long an action takes | | start_recording / stop_recording | Video recording of the mirrored screen | | list_targets / switch_target | Multi-device support |

Practical Evaluation Checklist

• [ ] macOS 15+ with iPhone Mirroring enabled
• [ ] iPhone connected via iPhone Mirroring
• [ ] Screen Recording + Accessibility permissions granted
• [ ] At least one MCP-compatible client installed (Claude Code, Cursor, Codex, Copilot)
• [ ] mirroir-mcp added to the client’s MCP configuration
• [ ] Ran a simple task: launch an app, tap a button, read screen text

Security Notes

• Mirroir requires Screen Recording and Accessibility permissions — both are full system-level access on macOS
• Only install from the official project (npm, Homebrew tap, or GitHub) — verify the install script URL
• The iPhone Mirroring connection is local; no cloud dependency for core functionality
• Community-contributed skills are available at github.com/jfarcand/mirroir-skills — review before running any third-party automation flows

FAQ

Q: Does this work on Windows or Linux? A: No. Mirroir is built on macOS iPhone Mirroring, which is a macOS-only feature. It will not work on Windows or Linux.

Q: Do I need a developer account? A: No developer account is required. iPhone Mirroring is a consumer feature built into iOS 17+ and macOS 15+.

Q: Can multiple iPhones be connected at once? A: Yes — the list_targets and switch_target tools let you manage multiple mirrored iPhones.

Q: Is the iPhone jailbroken? A: No jailbreak required. It uses the official macOS iPhone Mirroring framework.

Q: Does the AI see the actual screen or just a description? A: The describe_screen tool returns both a screenshot (base64 PNG) and a structured list of UI elements with coordinates. The screenshot tool returns the raw image.

Q: What MCP clients are officially supported? A: Claude Code, Cursor, GitHub Copilot (VS Code), and OpenAI Codex. The project is MCP-compliant so other MCP clients may work as well.

Conclusion

Mirroir solves the “AI agent can’t interact with mobile apps” problem by bridging to a real iPhone through macOS iPhone Mirroring. With 33 tools covering every input modality — taps, swipes, typing, gestures, and screen recording — it gives agents the same control a human user would have. If you are building AI agents that need to automate iOS app interactions, this is currently the only open-source MCP server purpose-built for that task.