Developer API

Quickstart

Create a key under Profile - API keys (shown once, store it securely), then:

curl -X POST https://audit.domains/api/v1/appraise \
  -H "Authorization: Bearer ad_live_YOUR_KEY" \
  -H "Content-Type: application/json" \
  -d '{"domain": "example.com"}'

Keys are sent as Authorization: Bearer ad_live_... or x-api-key. Requests without a key receive 401 key_required; keys on the free plan receive 402 paid_plan_required.

A successful appraisal returns this shape (values illustrative; fields are contract-locked and the interval always accompanies the point figure):

{
  "domain": "example.com",
  "valueUSD": 12400,
  "valueRange": { "low": 7400, "mid": 12400, "high": 20900 },
  "tier": "mid-market",
  "confidence": 62,
  "factors": [
    { "key": "brandability", "label": "Brandability", "score": 61 },
    { "key": "length", "label": "Length", "score": 74 }
  ],
  "methodologyVersion": "4.14",
  "engineVersion": "20.0.0",
  "comps": [
    { "domain": "shop.com", "priceUsd": 3500000, "year": "2014",
      "venue": "Sedo", "similarity": 82,
      "basis": "Shared keyword" }
  ],
  "appraisalHash": "8c1f...e2a9",
  "inputCommitment": "b44d...91c7",
  "resultDigest": "f01a...77de",
  "generatedAt": "2026-08-05T12:00:00.000Z",
  "dataEdition": "2026-07-01"
}

Endpoints

HEAD /api/v1/comps behaves as GET without a body and still spends 1 API call - do not use it as an uptime probe; poll GET /api/v1/usage or the status page instead. On /api/v1/bulk, each failed domain comes back with its own error code: appraisal_failed (the reserved call is refunded) or quota_exhausted (that domain was never charged).

Every response that reached the meter carries X-Credits-Limit, X-Credits-Remaining and X-Credits-Reset headers reporting your monthly API allowance and the date it refills - that is any 200, and the 402 you get when the allowance is spent. A request rejected before metering carries no meter to report: a malformed body, a missing or unknown key, or a 429 from the per-address limit. Read the meter without spending a call at GET /api/v1/usage.

X-RateLimit-Limit and X-RateLimit-Remaining currently repeat the same monthly allowance figures. Those names conventionally describe a short-window request limiter, which this API now has separately, so they will be redefined to report that limiter instead - not before an announced notice period of at least 30 days. Read X-Credits-* for the allowance meter.

Scopes

Each key is created with a scope. A read key can call the endpoints that spend no credits; a write key can call every endpoint. An endpoint that spends an API call refuses a read key with 403 insufficient_scope, and the scope is checked before the credit is debited, so a refused call costs nothing.

Issue read-scoped keys to anything that only searches recorded sales or polls the allowance at GET /api/v1/usage, so a leaked key from that integration cannot spend the account's monthly API allowance.

Request limits

Each key has a request limit per 60-second window, applied per key rather than per IP address, so several servers sharing one key share one budget and one server behind a shared address is not affected by its neighbours.

Over the limit, calls return 429 rate_limited with a Retry-After header in seconds and an X-RateLimit-Reset header giving the Unix second the window ends. Retry after that point rather than immediately.

Plans and credits

Each paid plan includes a monthly API allowance alongside its site appraisal credits. The two are metered separately: spending one does not spend the other.

The free plan does not include API access - keys created on a free account return 402 paid_plan_required until the account upgrades.

When the monthly API allowance is spent, calls return 402 api_quota_exhausted and site appraisal credits are unaffected. Wait for the monthly reset, or move to a larger plan on the pricing page.

Errors

Every failure returns the same JSON shape, and some codes add a field of their own on top of it - plan here. error is the only field to match on; it is stable and a given code always arrives with the same status. retryable says whether repeating the identical request can succeed, and requestId - also sent as the X-Request-Id header on every response, including successes - identifies the call in support requests.

{
  "plan": "core",
  "error": "api_quota_exhausted",
  "message": "This account's monthly API call allowance is spent.",
  "detail": "Wait for the monthly reset, or move to a larger plan at https://audit.domains/pricing. Site appraisal credits are a separate allowance and are unaffected.",
  "requestId": "8f2a1c4e9b7d0000-LHR",
  "docsUrl": "https://audit.domains/developers#api_quota_exhausted",
  "retryable": false
}

Free interactive access for AI assistants

For one interactive question at a time, with no key or API credit, the site runs a hosted MCP server at https://audit.domains/mcp exposing appraise_domain, compare_domains, find_comparable_sales, search_recorded_sales and run_bracket. Recorded-sales search is a citation sample capped at five rows per call. Full row-level research belongs to the paid Sales Research workspace or API. It is rate limited and capped per request; programmatic and bulk volume belongs here on the keyed API. A GET on that URL returns the discovery document with the published limits.

claude mcp add --transport http audit-domains https://audit.domains/mcp

Verifiable by design

Every appraisal response includes its ODVS hashes (appraisalHash, inputCommitment, resultDigest) under the Open Domain Valuation Standard: the same domain on the same engine version always yields the same commitment, and a replay that produces different numbers is provably a determinism violation. The canonicalization scheme is published in the repo's ODVS spec and covered by an independent conformance suite.