Connect Vynix to Cursor

Why connect Cursor to Vynix

Cursor is great at editing code once it knows exactly what to change. The hard part is telling it. A screenshot and "the header looks off on mobile" sends the agent guessing at selectors and viewports. Vynix removes that gap: every annotation already carries the resolved CSS selector, the DOM snippet, the computed styles, the viewport, and any console errors or failed requests. Over MCP, Cursor reads that context on demand and edits the right component the first time.

Before you begin

• The Vynix widget installed on the site you are getting feedback on (one script tag, see the install guide).
• At least one annotation captured in a Vynix project.
• Cursor updated to a version with MCP support.
• Node.js 18 or newer on your machine (the MCP server runs through npx).

Get your Vynix API token

The MCP server authenticates as you with a personal API token. Open your Vynix account, create an API token, and copy it. The token is long-lived and revocable, so you can roll it at any time without touching your password.

Add the Vynix MCP server to Cursor

Cursor reads MCP servers from a JSON file: use a project-scoped .cursor/mcp.json to share the setup with your repo, or the global ~/.cursor/mcp.json for all projects. Add the Vynix server and paste your token.

{
  "mcpServers": {
    "vynix": {
      "command": "npx",
      "args": ["-y", "@usevynix/mcp-server"],
      "env": { "VYNIX_API_TOKEN": "your-token-here" }
    }
  }
}

Reload Cursor. In Settings, MCP, you should see the vynix server connect and list its tools. If it stays disconnected, check that Node is on your PATH and that the token has no stray spaces.

What Cursor can do with your feedback

• List your projects and open annotations, so you can ask "what feedback do I have on the marketing site?"
• Read one annotation in full: page, element, selector, XPath, computed styles, DOM snippet, and runtime errors.
• View the region screenshots attached to a note.
• Generate a ready-to-paste coding prompt, or run an AI diagnosis for ranked root causes and likely files.
• Mark an annotation in progress or done, add a comment, or open a GitHub issue from it.

A typical loop in Cursor

Ask Cursor "what feedback is open on this project?" and it lists your notes. Say "fix annotation 3" and it reads the captured element and errors, edits the exact component, and can mark the note resolved and drop a comment recording what changed. The dashboard stays in sync, so your reviewers see progress without a status meeting.