Status

Availability objective: 99.5 % — an objective, not a measured figure; measured uptime appears here once 30 days of history exist

What this page is: the API's own status document, read on the server when this page rendered (2026-10-08 02:44 UTC).

Surfaces

Every row on this page is computed by the API about itself: a self-check of what each surface needs (no test receipt is minted to check it, and the outside settlement service for pay-per-call is not checked). The only outside measurement is the uptime figure below, when it is shown.

The public log

The witness vouches for the log's consistency, not for the truth of any receipt. A timestamp bounds time and backdates nothing.

Measured uptime

Measured uptime is not shown right now: the outside monitor's figures are not connected to this page, or its read did not answer.

What the states mean

• A surface is degraded after one failed self-check and down after two in a row with no pass between. A server error on the surface within five minutes can mark it degraded, never down. Unknown means it is not enabled or cannot be checked here.
• The log rows judge work left waiting, not how recently anything happened: a quiet log is healthy. Records waiting to be added are degraded after 2 h and down after 3 h; a signature older than 26 h (a missed daily re-sign) is degraded, and older than 52 h is down.
• A checkpoint left without the witness's cosignature for 2 h, a log state left without a calendar for 2 h, or one left without its Bitcoin confirmation for 24 h is degraded. These rows are never down: an outside party going quiet does not take this service down.
• The write queue is degraded at 1,000 waiting or when a record has waited 5 min past its turn, and down after 1 h.

This read: computed by the API at 2026-10-08 02:44 UTC. The API reuses one computation for at most two minutes, and this page reuses its read for at most 30 seconds.