Documentation

Everything on this page is in the product today. The API reference has every endpoint; this page is the short version.

Getting started

  1. Create an account (14-day trial, no card). Verify your email: scans are gated until you do.
  2. Add the domain you want to measure and, optionally, the competitors you want to be compared against. Give the domain a display name; brand mentions are matched on it, not only on the domain.
  3. Run a scan. Each prompt is sent to the engines you pick (ChatGPT, Claude, Gemini, Perplexity) and every answer is parsed for citations and mentions of your domains and competitors.
  4. Read the results (next section), then schedule the scan daily or weekly from the domain page.

Reading results

  • Organic share is the headline: of the citations in answers to prompts that do not name you, how many were you. Branded prompts (which name you) are reported separately as brand echo, because a model asked about you will mention you.
  • Cited means the answer linked or named your URL or domain; mentioned means it named the brand without a link. Both are stored with the matched text and its position, so any count can be checked against the raw answer.
  • Cited instead lists the domains that took the citations you did not get, per engine.
  • Sentiment is scored for citations of your brand and your tracked competitors only, and answers to leading prompts are excluded. The brand score counts brand citations only.
  • A single scan is one sample on one day. Treat movement between two scans as noise until you have a run of them.

REST API

Create an API key in Settings and send it as Authorization: Bearer <key>. Responses are JSON; errors follow RFC 9457 with a request id you can quote. Lists are cursor-paginated. Full API reference.

MCP server

MCP-aware clients (Claude, Cursor and others) can query your data directly. The endpoint is https://cloud-alloy.co.uk/mcp (Streamable HTTP, JSON-RPC over POST) and it takes the same Authorization: Bearer <API key> header as the REST API. Three tools are exposed:

  • check-citation — is a domain cited or mentioned for a prompt, per engine
  • get-attribution-report — the attribution report for one of your domains
  • list-scans — your recent scans and their status

Discovery metadata is served at https://cloud-alloy.co.uk/.well-known/mcp, and every public page carries a mcp-server meta tag pointing at the endpoint.

Webhooks

Register an HTTPS endpoint and receive a signed POST for scan.completed, scan.failed and accessibility_scan.completed. Deliveries carry an HMAC-SHA256 signature in X-Geo-Signature-256; failed deliveries are retried and every attempt is listed. Endpoints and payloads.

Exports and shared reports

  • GET /api/v1/scans/{id}/export downloads a scan's results as CSV; GET /api/v1/scans/{id}/citations/export downloads its citations.
  • A scan can be published as a read-only shared report at a link; readers need no account, and the report carries the same method footer as the app.

Plans and tokens

Usage is metered in tokens: a scan costs a fixed number of tokens per engine, and analyses (sentiment, accessibility, optimisation) have their own prices, all visible on the pricing page and in Settings. The trial starts with a token allowance and no card; adding a payment method before it ends continues on Pro, otherwise the account pauses with its data kept.

Help

Signed-in users can report a bug from the app menu; the report captures the request id and recent client events so it can be traced. Service health is on the status page, planned work on the roadmap and releases in the changelog.