Tools

The @cite42/mcp server exposes seven tools, one per metered endpoint, mapping one-to-one to the REST API. You don’t call them by hand; describe what you want and your agent picks the tool.

The tools

The server exposes seven tools, one per metered endpoint. You don’t invoke them directly; you ask your agent in plain English, it picks the right tool, and it answers you in plain English from the JSON the tool returns. You never handle the raw response yourself. Each tool has its own page with an example prompt and the JSON it hands back:

Search: ask any question across the models. · Rankings: how brands rank in AI answers. · Compare: head-to-head vs a competitor set. · Citations: which sources the models cite. · Sentiment: how the models talk about a brand. · Keywords: search volume and related ideas. · Trends: Google Trends interest over time.

Reading your balance (/v1/usage) is a read-only call, not a tool; check it from the dashboard or over REST.

Choosing models

The five AI-model tools (search, rankings, compare, citations, sentiment) take an optional models list, so you can say “just ChatGPT” and pay for one model instead of four. Omit it and the tool queries all four. Keywords and trends are flat-priced data tools, so models doesn’t apply. See Per-model pricing.

Billing

MCP calls bill exactly like REST: per model queried, from the same prepaid balance. No agent-seat fee, no MCP markup. Each tool result includes the cost and remaining balance. See Credits & billing.