Pro feature

MCP Setup

VoiceDOM captures the DOM element + voice note you leave in the browser. The voicedom-mcp server hands those annotations to your AI agent over MCP, so it knows exactly which element you meant and what you asked for. Nothing to install — your MCP client runs it on demand via npx.

Claude Code

One command:

claude mcp add voicedom -- npx -y voicedom-mcp

Or add to ~/.claude.json:

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

Cursor / VS Code

Add to .cursor/mcp.json (project) or ~/.cursor/mcp.json (global):

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

OpenAI Codex CLI

Add to ~/.codex/config.toml:

[mcp_servers.voicedom]
command = "npx"
args = ["-y", "voicedom-mcp"]

Usage

Then ask your agent to call read_annotations — it returns every annotation (CSS selector, XPath, page URL, title, and your transcribed note). Pass a hostname (e.g. "example.com") to scope it to one site.

# In Claude Code, Cursor, or Codex, just say:
Read my VoiceDOM annotations for example.com and apply the changes.

Don't have VoiceDOM yet?

The MCP server is included in Pro — $19 lifetime, no subscription.

Add to Chrome