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 +-------------+
npx ccphoto --setup # One-time: register CCPhoto as an MCP serverThen in Claude Code, just say "take a photo" -- Claude handles the rest.
- Claude calls the
capture_photoMCP tool. - A QR code appears in your terminal.
- Scan it with your phone -- a camera page opens in your browser.
- 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.
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.
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.
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.
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.
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.
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).
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.
npx ccphoto --setupnpm install -g ccphoto
ccphoto --setupgit clone https://github.com/stepankaiser/ccphoto.git
cd ccphoto
npm install
npm run build
node dist/index.js --setupIf you prefer to register the MCP server manually instead of using --setup:
claude mcp add ccphoto -- npx ccphoto --mcp| Tool | Description | Parameters |
|---|---|---|
capture_photo |
Start the server and display a QR code. If a phone is already connected, it receives a push notification instead. | -- |
wait_for_photo |
Block until a photo is uploaded or the timeout expires. Call this immediately after capture_photo. |
timeout_seconds (optional, default: 120) |
get_latest_photo |
Return the most recent photo, or a specific photo by filename. | filename (optional) |
list_photos |
List all captured photos with filenames, timestamps, and sizes. | -- |
send_to_phone |
Send text, images, or rich generative UI to the phone display. The phone becomes a reference screen for instructions, diagrams, or pinouts. | text, image_base64, image_mime_type, speak, ui_spec (all optional) |
start_livestream |
Start a live video stream from the phone camera. Generates HTTPS certificate and returns QR code. | -- |
get_live_frame |
Get the latest frame from the live camera stream with freshness timestamp. | -- |
CCPhoto can also run independently, without Claude Code:
npx ccphoto # Start with defaults
npx ccphoto --port 4000 --output-dir ./photos # Custom port and directoryA QR code is printed to the terminal. Scan it, take photos, and they are saved to disk. The server stays running for multiple captures.
| Option | Description | Default |
|---|---|---|
--setup |
Register CCPhoto as an MCP server in Claude Code | -- |
--mcp |
Run as an MCP server (used internally by Claude Code) | -- |
--port <number> |
HTTP server port | 3847 |
--output-dir <path> |
Directory where photos are saved | ~/.ccphoto/captures/ |
--help |
Show help | -- |
| Environment Variable | Description | Default |
|---|---|---|
CCPHOTO_DIR |
Override the default photo storage directory | ~/.ccphoto/captures/ |
- 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.timingSafeEqualto 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.
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.
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.
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).
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.
npm install # Install dependencies
npm run build # Compile TypeScript
npm 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
- 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.
Contributions are welcome.
- Fork the repository
- Create a feature branch (
git checkout -b my-feature) - Make your changes
- Run
npm testto verify nothing is broken - Commit your changes (
git commit -m "Add my feature") - Push to your branch (
git push origin my-feature) - Open a pull request
When reporting a bug, please include your Node.js version, OS, and the steps to reproduce the issue.