CCPhoto — Phone Camera Bridge

CCPhoto

[npm version] [License: MIT] [Node.js]

Bridge your phone camera to Claude Code over the local network.

CCPhoto is an MCP server that lets you take photos on your phone and send them directly to Claude Code. Point your phone at a circuit board, a whiteboard, a label -- anything -- and Claude sees it instantly. No app install required; your phone's browser camera does all the work.

 +----------+        WiFi        +-----------+       MCP        +-------------+ |  Phone   |  photos, voice,   |  CCPhoto  |   image/text     |  Claude     | |  Camera  |  mode switches    |  Server   |   data            |  Code       | |  + Mic   | <--------------- |  (local)  | <--------------- |             | +----------+   TTS, guidance   +-----------+   send_to_phone  +-------------+

Quick Start

npx ccphoto --setup        # One-time: register CCPhoto as an MCP server

Then in Claude Code, just say "take a photo" -- Claude handles the rest.

How It Works

1. Claude calls the capture_photo MCP tool.
2. A QR code appears in your terminal.
3. Scan it with your phone -- a camera page opens in your browser.
4. Take a photo -- Claude receives the image directly.

No app install needed. The phone's native camera is accessed through the <input capture> HTML attribute.

Always-On Mode

The server starts on the first capture_photo call and stays running for the entire session. Your phone stays connected -- there is no need to re-scan the QR code. Take as many photos as you like, and retrieve any of them by asking Claude.

Photo Request Notifications

When your phone is already connected and Claude calls capture_photo again, the phone receives an SSE push notification. The capture button pulses to let you know Claude is waiting for a new photo. Just tap and shoot.

Bidirectional Messaging

Claude can send content to your phone using the send_to_phone tool. Text, markdown, and images appear in a collapsible panel on the phone. This turns your phone into a reference screen -- Claude can display wiring instructions, pinout diagrams, or code snippets while you work with both hands.

Photo Annotations

After taking a photo, an annotation screen appears where you can draw on the image before sending it. Use colored pens (red, blue, green, white) to circle components, draw arrows, or highlight areas of interest. Tap "Send" to upload the annotated image, or "Skip" to send the original photo without annotations.

Live Video Assistant

Start a semi-real-time video stream from your phone camera. Claude watches the live feed (one frame every 3 seconds) and provides guidance in real-time -- point the camera where Claude asks, and receive instructions on your phone screen.

Requires HTTPS (for camera access). On Android Chrome, accept the certificate warning once. Currently Android/Chrome only; iOS support planned.

Voice Interaction

Tap the microphone button on your phone to speak to Claude. Speech is transcribed using the browser's built-in Speech Recognition API and sent as text. Claude can respond with spoken audio using send_to_phone with speak: true -- the phone speaks the response aloud via Text-to-Speech.

This enables hands-free workflows: ask a question while holding a soldering iron, and hear the answer without looking at a screen. The mic button appears only on browsers that support Speech Recognition (Android Chrome).

Generative UI

Claude can generate rich, contextual UI on your phone based on what it sees. Point your camera at a product, component, recipe, or anything -- Claude analyzes it and sends a custom interactive interface to your phone display.

12 component types: Card, Text, Image, Metric, MetricGrid, Checklist, StepByStep, Alert, Table, Timer, Badge, and Divider. Claude composes these into contextual layouts -- a plant care card with watering checklist, a circuit component table with pin diagrams, or a step-by-step repair guide with timers.

Use send_to_phone with the ui_spec parameter to generate custom UI.

Installation

npx (recommended, no install)

npx ccphoto --setup

Global install

npm install -g ccphotoccphoto --setup

From source

git clone https://github.com/stepankaiser/ccphoto.gitcd ccphotonpm installnpm run buildnode dist/index.js --setup

If you prefer to register the MCP server manually instead of using --setup:

claude mcp add ccphoto -- npx ccphoto --mcp

MCP Tools Reference

Standalone CLI Mode

CCPhoto can also run independently, without Claude Code:

npx ccphoto                                  # Start with defaultsnpx ccphoto --port 4000 --output-dir ./photos  # Custom port and directory

A QR code is printed to the terminal. Scan it, take photos, and they are saved to disk. The server stays running for multiple captures.

CLI Options

Configuration

Security

• Session token -- A cryptographically random 32-character hex token is generated for each server session. Every request must include the token.
• Timing-safe comparison -- Token validation uses crypto.timingSafeEqual to prevent timing attacks.
• Local network only -- The server binds to your local network IP. No data leaves your network.
• No external services -- Photos are transferred directly between your phone and the server over HTTP on your LAN. Nothing is uploaded to the cloud.

Troubleshooting

Could not detect local network IP

Make sure your computer is connected to WiFi (not just Ethernet in some configurations). CCPhoto needs a LAN IP address that your phone can reach.

QR code does not scan

Try increasing your terminal font size or reducing the terminal width so the QR code renders at a scannable resolution. You can also copy the URL printed below the QR code and type it into your phone's browser manually.

Phone camera does not open

The camera capture feature requires HTTPS on some browsers, but most mobile browsers allow it on plain HTTP for local network addresses. If your browser blocks camera access, try using Safari (iOS) or Chrome (Android).

Upload fails

Verify that your phone and computer are on the same WiFi network. Corporate networks with client isolation may block device-to-device traffic. Try a personal hotspot or home network instead.

Development

npm install          # Install dependenciesnpm run build        # Compile TypeScriptnpm run dev          # Watch mode (recompile on change)npm test             # Run the test suite (94 tests)

Tests use the Node.js built-in test runner and require the tsx dev dependency for TypeScript execution.

Project structure:

src/  index.ts        CLI entry point and arg parsing  mcp.ts          MCP server with tool definitions  server.ts       HTTP server, SSE, upload handling  mobile-page.ts  Self-contained HTML for the phone camera page  storage.ts      Photo file management  network.ts      Local IP detection  token.ts        Session token generation and validation  types.ts        Shared TypeScript interfaces

Roadmap

• Real domain + Let's Encrypt certs -- Use a registered domain with valid certificates for zero browser warnings on any device, including iOS Safari.
• iOS live video support -- Live video streaming currently works on Android/Chrome. iOS Safari support requires trusted certificates.
• AR overlays on camera feed -- Draw annotations directly on the live camera view in real-time.
• Multi-camera support -- Connect multiple phones simultaneously for different viewing angles.
• WebRTC streaming -- Replace HTTP frame polling with WebRTC when Claude's vision processing latency improves.

Contributing

Contributions are welcome.

1. Fork the repository
2. Create a feature branch (git checkout -b my-feature)
3. Make your changes
4. Run npm test to verify nothing is broken
5. Commit your changes (git commit -m "Add my feature")
6. Push to your branch (git push origin my-feature)
7. Open a pull request

When reporting a bug, please include your Node.js version, OS, and the steps to reproduce the issue.

License

MIT