Re-check a page against an earlier receipt

Last updated: 2026-10-15

POST /v1/recheck fetches the page named in an earlier receipt again and tells you whether its text changed: unchanged, changed, gone or unreadable, the fields that differ, and a signed receipt naming both the old and the new hashes. Price: US$0.003 per call.

When to use it

• For page change detection without storing anything: keep the receipt, recheck when you need to know.
• Before reusing a quote or a citation you checked days ago.
• To keep a dated chain of evidence that a page stayed the same (each receipt names the one before).

Request

Send exactly one of the two baseline forms. The receipt from an earlier call:

{ "receipt": "eyJhbGciOiJFZERTQSIs…", "return_text": false }

Or a URL and the content_sha256 from our own fetch of it:

{ "url": "https://www.iana.org/help/example-domains", "content_sha256": "aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa", "return_text": false }

The page fetched is the baseline's url (the URL first asked for, not the final URL), so redirects are followed again. The body may be up to 168 KB, so a long receipt fits.

History (timeline)

Send your own earlier receipts for the same URL in history and the answer gains a timeline: each receipt in order, then this check. changed on each entry compares its clean-text hash with the entry before it (null for the first, or when a hash is missing). A receipt that does not verify, names another URL or is out of order is a 400 baseline_invalid before payment, with its index. The timeline is also in the signed recheck receipt. Requests without history are unchanged.

Answer

From the free sample (GET /v1/sample/recheck), shortened:

{
  "change": "unchanged",
  "changed": false,
  "differences": [],
  "baseline": {
    "source": "caller",
    "receipt_sha256": null,
    "url": "https://s2h6qt73jtzlgbgnnfacln2poq.srv.us/v1/sample/page",
    "content_sha256": "0e429607…"
  },
  "elapsed_seconds": null,
  "page": {
    "final_url": "https://s2h6qt73jtzlgbgnnfacln2poq.srv.us/v1/sample/page",
    "http_status": 200,
    "content_kind": "real",
    "content_sha256": "0e429607…",
    "raw_sha256": "179b9ac4…"
  },
  "text": null,
  "injection_flags": [],
  "advice": "The page's text is the same as in the baseline (byte for byte on the clean text). Markup or other bytes may still have moved; see differences.",
  "receipt": "eyJhbGciOiJFZERTQSIs…",
  "tier": "paid",
  "rail": "x402",
  "network": "eip155:8453"
}

• Receipts are signed by iss https://still-rapids-9yt7.here.now (keys at /.well-known/jwks.json), not by the origin you called. Check one free with POST /v1/receipt/verify (docs).
• differences lists every compared field that differs, whatever the verdict: content_sha256, raw_sha256, http_status, final_url, content_kind, and with a receipt baseline also server_ip and tls.cert_sha256 when both sides have them.
• baseline is what was compared against. With the url form, source is caller and the fields only a receipt carries are null.
• elapsed_seconds is the time since the baseline was fetched (null with the url form).
• page, text and injection_flags are the new fetch, as [/v1/fetch](/docs/fetch) gives them.

What "changed" means

The comparison is byte for byte on the clean text we extract (content_sha256). Markup-only changes are not changes: raw_sha256 may differ while the answer is unchanged. A page with a date, a counter or rotating text reads as changed on every visit. A content_sha256 from another tool will not match ours: use the hash from our own fetch.

Quote

With quote, the quote is matched on the live page again and the answer adds quote, with a verdict:

With quotes (a verify/quotes receipt), each quote gets its own verdict in quotes, in the order sent: index, quote_sha256, verdict (the same values as above), was, now, score, start, end, occurrences, and error when that quote was too costly to match now (its verdict is then unchecked). If every quote is too costly, the call is a 422 match_too_costly, not charged.

Chaining

The recheck receipt is itself a valid baseline: send it to the next call. Its result.baseline.receipt_sha256 is the SHA-256 of the receipt you sent, so a chain can be checked offline with verify-receipt.mjs.

Charges and errors

The call is charged like /v1/fetch whenever it answers, gone and unreadable included: the fetch was done and the receipt is evidence of it. If we read no page at all (a timeout, a connect error, a bad decode, too many redirects or a redirect we refuse), the call answers 502 upstream_unreachable with details.reason set, and is not charged. Every baseline problem is a 400 before payment, never charged:

• bad_request: neither or both baseline forms, or content_sha256 without url.
• baseline_invalid: the receipt does not verify against this service's keys, is not a fetch, verify/quote, verify/quotes or recheck receipt, is a sample, or names no URL; details.reason says which.

The other errors are those of /v1/fetch; see Errors. On the edge, a PDF page answers 501 not_on_edge, uncharged.

Compared with a hosted watch

No schedule, no webhook, nothing stored: you call when you want to know, and pay US$0.003 per check.