Getting started

Seven signals, all of them counts over things that actually happened. No estimated volumes, no model scoring another model on a 1–10 scale, no number without the evidence it came from.

An audit is the starting point: one domain in, an evidence-backed picture out. It takes a few minutes and you watch it run.

Want to see it before you sign up? The free scan runs the same engines and the same GEO check on a smaller sample — three generic prompts, one answer each — and says so on the result.

Ask an answer engine the same question twice and you can get two different brand lists. That single fact determines the whole design: we sample repeatedly and publish intervals instead of a point.

answers per poll = prompts (≤12) × engines × runs per prompt (5)

rate            = k mentions / n answers
interval        = Wilson 95% around that rate
delta vs 7d ago = shown only if a two-proportion test clears it

Every formula, plus the seven limitations we cannot engineer away, is on /methodology.

AuditYourBrand ships an MCP server: a stateless streamable-HTTP JSON-RPC endpoint exposing 21 read-only tools — 11 native tools plus 10 proxied classic-SEO tools. Auth is the same bearer token the dashboard uses.

curl -s -X POST "$AYB_API/login/" \
  -H 'Content-Type: application/json' \
  -d '{"username":"admin","password":"…"}'

# → {"token":"…","username":"admin"}

{
  "mcpServers": {
    "audityourbrand": {
      "type": "http",
      "url": "https://your-install.example.com/api/mcp/",
      "headers": { "Authorization": "Bearer <token>" }
    }
  }
}

MCP is the data; skills are the workflow. A tool description can say what a tool returns — it cannot say which tool to reach for, in what order, or when a number means nothing at all. Five skills carry that judgement, as plain markdown you can read, edit and audit.

npx skills add https://github.com/<org>/AuditYourBrand

Every skill enforces two rules. Anti-fabrication: if AuditYourBrand does not return a value, the agent writes unknown — and no bare percentage, ever, without its sample size and interval. Anti-spend: name the tool, the reason and the rough cost, then wait for a yes before any credit-consuming call.

A typical chain is ayb-brand-audit → ayb-search-vs-ai → ayb-fix-gaps, with the weekly review on a cadence.

The public API is read-only and lives at /api/v1/. It uses its own bearer tokens — deliberately not your dashboard session — with a read scope and a throttle of 120 requests per hour per token.

curl -s -X POST "$AYB_API/tokens/" \
  -H "Authorization: Bearer <dashboard-token>" \
  -H 'Content-Type: application/json' \
  -d '{"name":"reporting-script"}'

# → {"id":1,"name":"reporting-script","token":"ayb_…","preview":"ayb_…abcd", …}

curl -s "$AYB_API/v1/visibility/summary/?brand=1" \
  -H "Authorization: Bearer ayb_…"