# Install whoisbrowsing With Claude Code, Cursor or Codex

To install whoisbrowsing with an AI coding agent, paste `Add whoisbrowsing (live visitor view) to this project. Follow https://whoisbrowsing.com/SKILL.md exactly.` into Claude Code, Cursor or Codex. The agent creates a site with no sign-in, adds one script tag in the right file for your framework, and gives you a watch link and a claim link.

## The prompt

```
Add whoisbrowsing (live visitor view) to this project. Follow https://whoisbrowsing.com/SKILL.md exactly.
```

No account is needed to start. The site is created without sign-in and you claim it later with the claim link.

## What the agent does

1. Reads `https://whoisbrowsing.com/SKILL.md`, a single file with the steps, the tag and the checks.
2. Calls the API to create a site: `POST https://whoisbrowsing.com/api/agent/sites` with a body like `{"domains":["localhost:3000"]}`. The reply holds a site ID, the snippet, a watch link and a claim link.
3. Looks at your project to find the framework, then inserts the tag where that framework expects it. See the guides for [Next.js](/install/nextjs), [React](/install/react), [Vue](/install/vue), [SvelteKit](/install/svelte), [Astro](/install/astro), [Remix](/install/remix), [Nuxt](/install/nuxt) and [Angular](/install/angular).
4. Gives you the watch link (to see visitors live) and the claim link (to keep the site under your Google sign-in).

The result is the same one tag you would paste yourself:

```html
<script src="https://live.whoisbrowsing.com/s.js" data-site="YOUR_SITE_ID" defer></script>
```

## The MCP server

If you would rather give your agent tools than a prompt, whoisbrowsing runs an MCP server at `https://whoisbrowsing.com/mcp`. It offers three tools: `create_site`, `get_install_instructions` and `check_live`.

### Claude Code

Run this in your terminal. The syntax comes from the [Claude Code MCP docs](https://code.claude.com/docs/en/mcp):

```bash
claude mcp add --transport http whoisbrowsing https://whoisbrowsing.com/mcp
```

Then run `/mcp` inside Claude Code to confirm the connection.

### Cursor

Add the server to `.cursor/mcp.json` in your project, or to `~/.cursor/mcp.json` for all projects, as described in the [Cursor MCP docs](https://cursor.com/docs/context/mcp):

```json
{
  "mcpServers": {
    "whoisbrowsing": {
      "url": "https://whoisbrowsing.com/mcp"
    }
  }
}
```

### Codex

Use the CLI, per the [Codex MCP docs](https://learn.chatgpt.com/docs/extend/mcp?surface=cli):

```bash
codex mcp add whoisbrowsing --url https://whoisbrowsing.com/mcp
```

Or add it to `~/.codex/config.toml`:

```toml
[mcp_servers.whoisbrowsing]
url = "https://whoisbrowsing.com/mcp"
```

Run `codex mcp list` to see it.

## Verify

Do not take the agent's word for it. Check three things:

1. The diff shows one `<script>` tag with your site ID, in the layout or HTML file for your framework, and nothing else changed.
2. The site ID belongs to the address you will open. Localhost and production are separate websites, so a site made for `localhost:3000` will not show your live domain.
3. Open the watch link, then open your site in a Chrome Incognito or Safari private window. You should appear as a visitor on your real page. Your normal browser is never counted once you've opened the watch link in it. The `check_live` tool can confirm the same thing from the agent side.

If you use a Content-Security-Policy, allow `script-src https://live.whoisbrowsing.com` and `connect-src wss://live.whoisbrowsing.com`.

## If the agent gets stuck

| What happens | What to do |
|---|---|
| The agent says it cannot open web pages | Turn on its web access, or connect the MCP server above, which serves the same instructions as a tool |
| It picked the wrong file | Point it at your root layout or main HTML file, or follow your framework's guide by hand |
| It made a site for `localhost:3000` but you are watching production | Ask it to create a site for your production domain too; each website is separate |
| You see nothing in the private window | Check the page source for the tag, the site ID, and any CSP or ad blocker in the way |

## When an agent is not the right route

For a hosted site builder you edit in the browser, such as [WordPress](/install/wordpress) or [Shopify](/install/shopify), the agent has no project to edit. It can still create the site and hand you the tag, but you paste the tag yourself; the [install page](/install) lists every platform. You can also start from [/try](/try). To see what you are installing, read [how it works](/how-it-works) and the [privacy page](/privacy), or go back to the [home page](/).