Back to changelog
v0.43.0

MCP Server

Nvisy is now available to AI agents over the Model Context Protocol. Point any MCP-compatible client at @nvisy/mcp and redact documents, images, and audio from the conversation you are already in.

The official Model Context Protocol server is published as @nvisy/mcp. It exposes the platform to any MCP-compatible client, so redaction happens in the conversation rather than in a separate tab.

It is built on the TypeScript SDK and tracks its version, so the two stay in step.

Getting Started

Run it directly with npx:

NVISY_API_TOKEN=your-api-token npx @nvisy/mcp

For Claude Code:

claude mcp add nvisy --env NVISY_API_TOKEN=your-api-token -- npx -y @nvisy/mcp

For Claude Desktop, add the server to your configuration file:

{
  "mcpServers": {
    "nvisy": {
      "command": "npx",
      "args": ["-y", "@nvisy/mcp"],
      "env": {
        "NVISY_API_TOKEN": "your-api-token"
      }
    }
  }
}

Features

  • Redact documents, images, and audio without leaving the conversation
  • Inspect detections and adjust policies through natural language
  • Runs standalone over stdio, or embedded as a library
  • Built on the official Nvisy TypeScript SDK, and versioned alongside it
  • Full TypeScript support with strict typing throughout

Configuration

The server reads its settings from the environment:

Variable Required Default Description
NVISY_API_TOKEN Yes — API token used to authenticate against the Nvisy API
NVISY_BASE_URL No https://api.nvisy.com Base URL of the Nvisy API, for self-hosted deployments

Requires Node.js 24 or newer. The server is MIT licensed, and the source lives at nvisycom/mcp.