Use CoreViz as visual memory for Claude Code and other AI agents via the Model Context Protocol

MCP Server

The @coreviz/cli package includes a built-in MCP (Model Context Protocol) server that exposes your entire CoreViz visual library as tools for Claude Code and other MCP-compatible AI agents. This turns CoreViz into visual memory — your AI agent can browse, search, tag, and upload photos directly from chat.

What is MCP?

The Model Context Protocol is an open standard for giving AI assistants access to external tools and data sources. By running the CoreViz MCP server, Claude Code gains the ability to understand and work with your visual library without you having to copy-paste image URLs or describe your photos manually.

Setup

1. Log in to CoreViz

npx @coreviz/cli login

This stores your session token locally — the MCP server reads it automatically. No credentials in config files.

2. Connect to your MCP client

Claude Code — Install the plugin (recommended):

claude plugin marketplace add coreviz/cli
claude plugin install coreviz@coreviz

Or configure MCP manually in ~/.claude/settings.json:

{
  "mcpServers": {
    "coreviz": {
      "command": "npx",
      "args": ["coreviz-mcp"]
    }
  }
}

Claude Desktop — Add to ~/Library/Application Support/Claude/claude_desktop_config.json:

{
  "mcpServers": {
    "coreviz": {
      "command": "npx",
      "args": ["coreviz-mcp"]
    }
  }
}

Other MCP clients — Add to your project's .mcp.json:

{
  "mcpServers": {
    "coreviz": {
      "command": "npx",
      "args": ["coreviz-mcp"]
    }
  }
}

3. Verify the connection

In Claude Code, run /mcp and confirm that the coreviz server appears with all its tools listed.

Available Tools

Tool	Description
`list_collections`	List all collections in your workspace
`create_collection`	Create a new collection
`browse_media`	Navigate folders and list media items in a collection
`search_media`	Semantic search across all your media using natural language
`get_media`	Get full details, tags, and detected objects for a specific item
`get_tags`	Aggregate all tags across a collection
`find_similar`	Find visually similar images using an object ID
`analyze_image`	Run AI vision analysis on an image URL
`create_folder`	Create a folder inside a collection
`move_item`	Move a media item or folder
`rename_item`	Rename a media item or folder
`add_tag`	Add a tag to a media item
`remove_tag`	Remove a tag from a media item
`upload_media`	Upload a local photo or video file to a collection

Usage Examples

Once connected, you can interact with your visual library through natural language in Claude Code:

Search my library for photos of the product launch event

Upload all JPGs from ~/Downloads/shoot-2026 to my "Campaign 2026" collection

Tag all photos in the "Portraits" collection with category=headshot where a person is visible

Find images similar to the face in object ID abc123 from media item xyz456

Authentication Options

Run npx @coreviz/cli login once. The token is stored in your local config and used automatically.

Option 2 — Environment variable

Pass an API key via env when you don't want to use the device login flow:

{
  "mcpServers": {
    "coreviz": {
      "command": "npx",
      "args": ["coreviz-mcp"],
      "env": {
        "COREVIZ_API_KEY": "your_api_key_here"
      }
    }
  }
}

Local Development Override

When developing against a local CoreViz instance, override the base URL:

{
  "mcpServers": {
    "coreviz": {
      "command": "node",
      "args": ["/path/to/@coreviz/cli/bin/mcp.js"],
      "env": {
        "COREVIZ_API_URL": "http://localhost:3000"
      }
    }
  }
}

Next Steps

Library SDK: Use the same functionality programmatically in your own applications
Authentication: Learn more about authentication options

MCP Server

On this page