Skip to main content
The cloro MCP server exposes cloro’s sync monitor endpoints as tools for any MCP-capable client: Claude, Claude Code, Cursor, or your own agent. Once connected, you can ask your agent to “check whether ChatGPT recommends us for best SERP API” and it will call the cloro API and reason over the parsed result: the answer text and the sources it cited. The server is a thin wrapper over api.cloro.dev. It holds no business logic and no secrets: each request carries your own cloro API key, which it forwards as a Bearer token, so the API itself enforces authentication, credit billing, rate limiting, and concurrency limits. Tool input schemas are the same schemas the API validates with, so the MCP contract can’t drift from the API reference.

​
Prerequisites

  • An MCP-capable client (Claude Desktop, Claude Code, Cursor, or any other)
  • A cloro API key from the dashboard (see Authentication)

​
Connect your client

The hosted server speaks Streamable HTTP at https://mcp.cloro.dev. Pass your API key in the URL path so it works even in clients that can’t set custom headers:
If your client can set headers, you can instead point it at https://mcp.cloro.dev/mcp and send your key as Authorization: Bearer YOUR_CLORO_API_KEY.
In Claude Desktop and Cursor, add this block to the app’s MCP settings (Settings → Developer / MCP). In Claude Code, run claude mcp add --transport http cloro https://mcp.cloro.dev/YOUR_CLORO_API_KEY/mcp.

​
Available tools

The AI assistant tools (scrape_chatgpt, scrape_gemini, scrape_copilot, scrape_perplexity, and scrape_grok) require a prompt and a country (ISO 3166-1 alpha-2), and accept an optional state for US state-level targeting. scrape_google_ai_mode also requires a prompt and a country, but targets sub-country with location or uule rather than state, and accepts device. scrape_google_news requires a query and a country, and accepts device and pages. scrape_google runs in one of two modes. Either pass a query with a country (plus optional location or uule, device, and pages), or pass a complete google.com/search url, which carries the query, location, and pagination itself; in url mode country can be derived from the URL’s gl parameter. It also accepts include.aioverview to extract Google’s AI Overview and include.paaAioverview to hydrate the AI answers in People Also Ask. Every tool accepts an optional include object to add heavier payload fields; leave it unset for the leanest response. Use list_countries (optionally filtered by model) and list_states (with a country) to discover valid geo-targeting codes.
scrape_grok is registered, but Grok is temporarily unavailable. Calls to it fail until the provider is restored.

​
Credits and limits

Each tool call runs synchronously and consumes credits at the standard per-provider rates. See Providers and Billing & credits. Calls count against your concurrency limits.
For large batches, such as tracking many prompts across engines, use the async API directly instead of looping tool calls.

​
Example prompts

Once connected, ask your agent directly:
  • “Ask ChatGPT what the best SERP APIs are and list which vendors it cites.”
  • “Compare how Perplexity and Copilot answer ‘best CRM for startups’. Which sources do they cite?”
  • “Does Google’s AI Overview for ‘ai visibility tracking’ mention cloro? Check in the US and Germany.”
  • “Pull the latest Google News articles about our brand.”