Back to CMO.ie

MCP connector

Ask Claude about your AI visibility.

The CMO.ie MCP server lets Claude query your live CMO.ie data in natural language. Install it once and you can ask questions like “How did my AI visibility change this week, and what's the single biggest gap to act on?” in any Claude client that supports MCP.

Install

  1. Mint a REST API key at Settings → REST API keys. Select the scopes you want Claude to have access to. For full functionality, tick all six.
  2. Copy the plaintext key (shown once, starts with cmo_).
  3. In Claude.ai, Claude Desktop, Claude Code, or any other MCP client, add a new connector with these details:

Server URL

https://cmo.ie/api/mcp

Transport

streamable-http

Authentication

Authorization: Bearer cmo_…

Available tools

Claude automatically picks the right tool based on your question. The six read-only tools exposed to the model:

  • list_projectsList every project in your organisation. Claude calls this first when it doesn't know which project you mean.
  • get_visibilityVisibility %, share of voice, average position, sentiment distribution for a project over a window.
  • list_gapsRanked list of sources where competitors appear and you don't, with per-source playbooks.
  • list_sourcesDomains (or URLs) AI cites when answering your tracked prompts, with retrieval and citation rates.
  • get_prompt_detailFull visibility arc for one specific tracked prompt — per-model snapshot, sources, brands named, response history.
  • get_recent_chatsMost recent individual AI responses with snippets. Useful for pulling direct evidence.

Example prompts

  • “How did my AI visibility change in the last 30 days?”
  • “Which competitor is gaining the most share of voice on my prompts?”
  • “Give me the three highest-opportunity gaps I should act on this week, and what the playbook looks like for each.”
  • “What's the full history on the prompt about ‘best accountants in Dublin’ — did any AI mention us?”

REST equivalents

Every MCP tool is a thin wrapper over the REST API. Same data, same auth, same scope model — use whichever interface suits the tool you're building.