CameraMCPServer

# MCP Camera Service ## Feature Overview This project is a multi-protocol camera control service based on Node.js/TypeScript, strictly following the [Model Context Protocol (MCP)](https://github.com/modelcontextprotocol/modelcontextprotocol) TypeScript SDK. It supports: - Camera enumeration, photo taking, video recording (file/streaming), and more - MCP tool interfaces (listCameras, takePhoto, startVideo, stopVideo) - Multiple transport modes: stdio, HTTP, SSE - Support for Windows (dshow) and Linux (v4l2) - Support for concurrent video recording and streaming with multiple instances - Rich environment variable configuration, customizable logging - Video recording streaming supports streamUrl, clients can pull streams using httpx/curl/ffmpeg/ffplay --- ## Environment Dependencies - Node.js 16+ - ffmpeg (must be installed and available in PATH or provided by ffmpeg-static) - Recommended: Windows 10+/Linux, with a camera device --- ## Installation and Startup ```bash npm install # Configure the .env file (see below) npm run build npm start # or npm run start:http ``` --- ## .env Configuration Instructions When executing locally, please copy .env.sample to .env and configure it with the parameters you need. ```ini # HTTP server port PORT=3001 # MCP service name/version MCP_SERVER_NAME=mcp-camera-server MCP_SERVER_VERSION=1.0.0 # Log level and file LOG_LEVEL=info LOG_FILEPATH=./camera.log # Photo/video file prefix and directory PHOTO_PREFIX=image VIDEO_PREFIX=video PHOTO_DIR=photos VIDEO_DIR=videos # Video context auto-clean time (seconds) FINISHED_VIDEO_CLEAN_SPAN_SECOND=1800 # Stream URL timeout (milliseconds) STREAM_URL_EXPIRE_MS=300000 ``` --- ## Main MCP Tool Interfaces ### 1. listCameras - Enumerate Cameras Function Description: Lists all available camera devices Input Parameters: None Output Parameters: - cameras: List of cameras - cameraID: Camera identifier - name: Device name - description: Device description ### 2. takePhoto - Photo Taking Function Function Description: Captures a photo from the specified camera and saves it Input Parameters: - cameraID: Camera identifier - filepath: Path to save the photo Output Parameters: - success: Whether the operation was successful - imagePath: Path where the photo is saved (returned on success) - error: Error message (returned on failure) ### 3. startVideo - Start Recording/Streaming Function Description: Starts video recording or live streaming Input Parameters: - cameraID: Camera identifier - filepath: Path to save the video file (optional, if not provided, streaming mode is activated) - duration: Duration of the recording (milliseconds, optional) Output Parameters (file mode): - success: Whether the operation was successful - videoPath: Path to the video file - videoID: Recording session ID Output Parameters (streaming mode): - success: Whether the operation was successful - videoID: Recording session ID - streamUrl: Streaming access address ### 4. stopVideo - Stop Recording/Streaming Function Description: Stops the ongoing recording or streaming Input Parameters: - videoID: Recording session ID (returned by startVideo) Output Parameters: - success: Whether the operation was successful - message: Operation result message - videoPath: Path to the video file (exists in file mode) - error: Error message (returned on failure) --- ## MCP Tool Interface Testing Method - After downloading this repository, navigate to the root directory `cd myCameraNodejs` - Use `npm link .` to link it as a global command - Please start modelcontextprotocol/inspector for testing using `npx modelcontextprotocol/inspector`:  In the interface: - `Transport Type`: Select `STDIO` - `Command`: Select `npx` - `Arguments`: Enter `mcp-camera-server` Click the `Connect` button to connect to the camera service. If you start the service using `npx run start:sse`, fill in the `URL` as `http://127.0.0.1:3001/sse` (note that 3001 is specified by PORT in .env) ## Video Stream Testing Method ### 1. Save as ts file using curl ```bash curl http://localhost:3001/video/stream?videoID=xxxx-xxxx-xxxx -o output.ts ``` ### 2. Directly play using ffplay ```bash ffplay http://localhost:3001/video/stream?videoID=xxxx-xxxx-xxxx ``` ### 3. Save as mp4 file using ffmpeg #### Directly transcode to mp4 (recommended) ```bash ffmpeg -i "http://localhost:3001/video/stream?videoID=xxxx-xxxx-xxxx" -c:v libx264 -preset veryfast -crf 23 -f mp4 output.mp4 ``` #### Save as ts file only (no transcoding, fastest) ```bash ffmpeg -i "http://localhost:3001/video/stream?videoID=xxxx-xxxx-xxxx" -c copy output.ts ``` --- ## Common Issues and Troubleshooting - **ffplay/ffmpeg reports Invalid data found when processing input** - Please confirm that the server's streaming format is mpegts (fixed) - **streamUrl can only be consumed once/times out** - This is by design; you need to call startVideo again to get a new streamUrl - **Camera not found/photo taking failed** - Please first use listCameras to obtain the cameraID and ensure ffmpeg can recognize it - **Video file cannot be played** - It is recommended to use ffmpeg/ffplay for transcoding or playback - **Streaming timeout/recording context cleanup time** - Can be configured through .env with STREAM_URL_EXPIRE_MS and FINISHED_VIDEO_CLEAN_SPAN_SECOND --- ## Other Notes - Supports concurrent recording/streaming with multiple instances - Detailed logging for troubleshooting - Supports Windows/Linux, automatically adapts to dshow/v4l2 - Supports multiple modes of MCP stdio/http/sse For more questions or requests, feel free to open an issue or contact Jackgao19700@qq.com or gaoshengqiang19700@gmail.com It is certified by MCPHub. You can get it from https://mcphub.com/mcp-servers/jackgao19700/cameramcpserver.