Status

The PatentsView-compatible API at https://patentref.io/api/v1/. Figures refresh once a minute, the 30-day figures every 10 minutes; this page was generated at 2026-11-06T16:35:23Z. Machine-readable twin: status.json.

Coming change: Privacy Policy

Version beta-4 of the Privacy Policy takes effect on 2026-11-09. Until then the version before it applies. What is new: a waitlist you can join at patentref.io/waitlist/ (section 1, Waitlist, and section 5), which is in our nightly backups for up to 45 days after an address leaves it (section 5); weekly counts of visits per page and per referring site, taken from the web server's access log, which this version says also records the referring page (section 1, Website visits); a weekly count of sign-ups per campaign code, which is not stored with your account (the same paragraph); a correction: beta-3 called the copy of our backups on Google Drive encrypted, but we do not encrypt our backups ourselves (section 6 says how they are copied and protected); and hello@patentref.io as the contact address until our own mailboxes are live.

Coming change: Processors and data processing agreements

Version beta-4 of the Processors and data processing agreements takes effect on 2026-11-09. Until then the version before it applies.

Operational All checks passing. Code 1955e1c, data 20261103.2, health check on the box: ok.

100.000%

uptime since 2026-10-06

minutes down of 45,408 elapsed

7 ms

p95 latency, last 24 h

320

API requests, last 24 h

The uptime figure counts from 2026-10-06 03:47 UTC: the GitHub probe until 2026-10-15 02:57 UTC (0.0 minutes down), then UptimeRobot (0.0 minutes down).

External check

UptimeRobot checks /api/v1/meta/health/ from its own network (region EU) every minute and confirms a failure before it reports the API down. The box reads that record every 2 minutes. Since 2026-10-15 02:57 UTC its incidents are the down time of the uptime figure.

State: up since 2026-10-15T02:57:07Z
Last check: 2026-11-06T16:33:00Z (2 min ago), answered in 115 ms
Response-time points, last 24 h: 333 (points of UptimeRobot's response-time series, which groups its checks; not a count of checks)
Last incident: from 2026-10-15T02:52:31Z for 276 s: 404 Not Found (resolved)
Record read: 2026-11-06T16:34:06Z (1 min ago)

External uptime probe (GitHub)

GitHub Actions runs a probe every 5 minutes from GitHub's network against /api/v1/meta/health/ and one keyed query. The span between two consecutive failed probes counts as down; one failed probe between two passes counts as nothing. GitHub's scheduler can run a probe several minutes late or skip one under load, so the probe count varies. The history is a public commit log, mirrored to the box every few minutes.

Last probe: 2026-11-06T16:11:19Z (24 min ago), pass HTTP 200 in 596.2 ms, keyed query HTTP 200 in 608.7 ms, data 20261103.2
Last 24 h: 4 probes, 0 failed
Since clock start: 160 probes, 0 failed, 0.0 minutes down; measured from 2026-10-07T02:33:30Z, the 1366 minutes between the clock start and the first probe are unobserved and counted as up
Mirror age: 1 min

No outage recorded since the clock started.

Requests, errors, latency

From the per-request log on the box (every response under /api/v1/ except the account, health and status endpoints). Client errors are 4xx other than 403 and 429; rejected is 403 (no or bad key) and 429 (a limit); latency is measured over admitted requests, inside the app, without the network. Not served is the documented 501 of an endpoint whose data set is not in this release (/api/v1/meta/ lists them under deferred): it is not counted as a server error. Requests the service sends to check itself are not in these figures; they are counted on the last row.

Figure	last 24 h	last 30 days
Requests	320	1,652,231
Answered 2xx or 3xx	1	252
Client errors (4xx)	4	16,732
Rejected (403, 429)	315	1,635,242
Server errors (5xx)	0	5
Not served (documented 501)	0	0
Error rate (4xx plus 5xx, of all)	1.25%	1.013%
5xx rate	0.0%	0.0%
p50 latency	6.4 ms	1.7 ms
p95 latency	7.3 ms	7.9 ms
Slowest	7.3 ms	20562.3 ms
Account, health and status requests	3,044	65,415
Other paths (scanners, the root redirect)	694	13,733
The service's own checks (not in the figures above)	193	48,829

Data refresh

Serving: data version 20261103.2, built 2026-11-06T10:07:19Z, live since 2026-11-06T10:13:26Z
Last catalog check: 2026-11-04T06:18:16Z: new data, build started
Next build: none waiting
Checks of the live build: USPTO files fetched: pass; file formats unchanged: pass; database built: pass; database integrity: pass; row counts against the previous build: pass; the PatentsView API's example queries: pass; history-match score tested on past patents: pass; expiry audit of known patents: pass

The weekly check polls the USPTO catalog on Wednesdays at 06:10 UTC and starts a build only when a wanted file changed. A build never replaces the live file by itself.

Box

Resource	free	threshold
Disk, system	81.5%	ok
Disk, data	43.1%	ok
Memory available	96.0%	ok

Last backup: 2026-11-06T03:34:03Z, ok in 194 s (Drive ok, Storage Box ok)
Watchdog: last run 2026-11-06T16:34:06Z: api_local ok, api_public ok, api_unit ok, disk_root ok, disk_data ok, memory ok, refresh_units ok, refresh_build ok, backup ok, probe_mirror ok, registry_api ok, registry_db ok, registry_keys ok, uptimerobot ok

Alerts

No open alerts. Most recent line: 2026-11-06T10:17:39Z refresh_units recovered.

The on-box watchdog runs every 2 minutes: API and public health, disk and memory headroom, the refresh and backup timers. A breach alerts the operator (once per 24 hours per check) and adds a line here.