playbooks

Home / MCP / Webcam MCP Server

Webcam MCP Server

Provides webcam access with capture, screenshot, and sampling endpoints for MCP clients.

javascript
97stars
View on GitHub
Installation
Add the following to your MCP client configuration file.

Configuration

View docs
{
    "mcpServers": {
        "webcam_http": {
            "url": "http://localhost:3333/mcp"
        }
    }
}

You run mcp-webcam to expose live webcam access through an MCP (Multi-Client Protocol) server. It lets you capture images, take screenshots, and share a live resource with clients while supporting sampling requests from compatible MCP clients. This guide walks you through practical usage, installation, and configuration so you can run and use the server locally or in a cloud/deployed environment.

How to use

Start with the two main ways to run the server locally: as a stdio process via npx or as a Streaming HTTP server that serves the MCP endpoint at a dedicated URL.

How to install

Prerequisites: You need Node.js installed on your platform. You will use npm and npx to run the MCP server. Ensure your environment has access to a network port for the UI and MCP endpoint (default 3333). Then you can install and run the server using one of the following approaches.

Install and run in STDIO mode using NPX

npx @llmindset/mcp-webcam

Or run in Streaming HTTP mode so the MCP Server is available at a dedicated URL while the UI remains at the local port.

npx @llmindset/mcp-webcam --streaming

If you prefer to run with Docker, you can start a container that exposes the UI and MCP endpoint on port 3333 by default. The container runs in streaming mode unless you set a transport mode to stdio.

docker run -p 3333:3333 ghcr.io/evalstate/mcp-webcam:latest

Additional config and start options

To customize how you access the server, use environment variables when running with Docker or in your environment. Common options include the port, host binding, and public-facing host for MCP clients.

Available tools

capture

Capture an image from the webcam and expose it as a resource

screenshot

Take a screenshot from the current camera view and expose it as a resource

sampling

Send a sampling request to a connected MCP client with a user-provided message and receive results