Aurora MCP Server — Setup & Configuration

What is MCP?

The Model Context Protocol (MCP) is an open standard that lets AI assistants connect to external tools and data sources. Aurora's MCP server lets you query your GEO tracking data directly from your favorite AI client — no dashboards needed.

What You Get

Connect Aurora MCP and your AI assistant can:

• Track visibility over time — daily visibility trends for your brand and every tracked competitor
• Measure citation share — the share of AI answers that cite your domain (the same metric as the dashboard chart)
• Analyze share of voice — own-brand vs competitor mentions, overall and per AI engine
• Profile competitors — who wins which prompts, their citation rates, and the sources fueling them
• Audit sources — which domains and URLs AI models cite most, with source-type and freshness intel
• Read raw answers — the actual text ChatGPT, Perplexity, Gemini & co. produced for your prompts
• Use insights reports — executive summaries and prioritized recommended actions
• Track AI shopping — product win rates, rankings, and price accuracy in AI shopping answers
• Run built-in reports — one-command weekly pulse, engine scorecard, and competitor-movers workflows
• Manage prompts (optional write access) — create or pause tracked prompts and accept recommended actions
• Create campaigns (separate opt-in grant) — set up a new brand, competitors and prompts without leaving your AI client

See the full tool reference for details.

Tracked AI Providers

Aurora monitors your brand across these AI platforms:

Prerequisites

• An active Aurora Intelligence subscription
• At least one campaign with tracking data
• An MCP-compatible client (see setup instructions below)

Setup

Server URL (all clients):

https://aurora-mcp.aurora-intelligence.workers.dev/mcp

Claude (Desktop & Web)

The easiest path is a custom connector — no config file needed:

1. Go to Settings → Connectors (on claude.ai or in Claude Desktop).
2. Click Add custom connector and enter:
   • Name: Aurora
   • URL: https://aurora-mcp.aurora-intelligence.workers.dev/mcp
3. Click Connect. A browser window opens where you sign in with your Aurora Intelligence account and click Authorize. If you also want Claude to be able to create/pause prompts, tick "Allow write access" on that screen — otherwise leave it unchecked for read-only.
4. Back in Claude, start a chat and try: "List my Aurora campaigns."

Claude Code (CLI)

1. Run this command in your terminal:

claude mcp add aurora --transport http https://aurora-mcp.aurora-intelligence.workers.dev/mcp

2. Connect via the /mcp command: Type /mcp in Claude Code and select the Aurora server. A browser window will open for you to sign in with your Aurora account.

ChatGPT

ChatGPT supports MCP servers as connectors (available on paid plans; custom connectors may require enabling Developer mode under Settings → Apps & Connectors → Advanced).

1. In ChatGPT, go to Settings → Apps & Connectors (called Connectors in some versions).
2. Click Create / Add custom connector and enter:
   • Name: Aurora
   • MCP server URL: https://aurora-mcp.aurora-intelligence.workers.dev/mcp
   • Authentication: OAuth
3. Save. ChatGPT opens a browser window where you sign in with your Aurora Intelligence account and click Authorize.
4. In a chat, enable the Aurora connector (via the tools/plus menu) and ask: "List my Aurora campaigns."

Cursor

1. Add to your Cursor MCP settings (.cursor/mcp.json in your project root):

{
  "mcpServers": {
    "aurora": {
      "url": "https://aurora-mcp.aurora-intelligence.workers.dev/mcp"
    }
  }
}

2. Restart Cursor. The first time you use an Aurora tool, a browser window will open for authentication.

Authentication

Aurora's MCP server uses OAuth 2.1 — the same secure standard used by Google, GitHub, and other platforms. Here's how it works:

1. You ask your AI assistant to use an Aurora tool (e.g. "List my campaigns")
2. Your AI client automatically opens a browser window
3. You sign in with your Aurora Intelligence account (the same email/password you use on aurora-intelligence.tech)
4. After signing in, the browser redirects back and your AI client is connected

No API keys, no tokens to copy, no secrets to manage. The token is stored by your AI client and lasts for 7 days. After that, you'll be asked to sign in again.

During sign-in you can optionally grant write access, which enables tools for creating/deactivating prompts and accepting insights actions. Read-only is the default.

A second, separate checkbox grants campaign creation. It is deliberately not part of write access: a new campaign starts recurring, billable tracking and counts against your plan's campaign limit, so it is opted into on its own. Granting write access alone does not allow it.

Personal Access Tokens (alternative)

If you prefer a long-lived credential — for headless setups, scripts, or to avoid the weekly re-login — create a Personal Access Token under Profile → Personal Access Tokens in the Aurora dashboard. Configure your MCP client to send it as a Bearer token:

{
  "mcpServers": {
    "aurora": {
      "url": "https://aurora-mcp.aurora-intelligence.workers.dev/mcp",
      "headers": {
        "Authorization": "Bearer aur_pat_..."
      }
    }
  }
}

Tokens can be revoked at any time from the same page — revocation takes effect immediately. Treat tokens like passwords: they grant access to all your campaign data.

Example Queries

Once connected, try asking your AI assistant:

• "List my Aurora campaigns"
• "How has my brand's visibility developed over the last 30 days?"
• "What's my citation share vs competitors this month?"
• "What's my share of voice per AI engine?"
• "Which competitor is winning the prompts I track, and with which sources?"
• "Show me what ChatGPT actually said about us last week"
• "Summarize my latest insights report and its top 3 recommended actions"
• "Which of my product pages win AI shopping recommendations?"

Or run a built-in prompt — your MCP client lists them next to the tools:

Want full workflow examples with sample responses? Check out the Use Cases page.