Live measurement is paused.
The newest measurement is older than the 5 minute freshness window, or the feed did not answer, so nothing on this page is shown as current. That is not by itself an outage.
last measured: operational 2026-10-16 06:36 UTC · probe every 60 seconds · a measurement older than 5 minutes reads paused, never operational
Surfaces
external probe · 1 minute interval
Web and console
pausedcovered by the control-plane probe; the current result is fetched in your browser · last measured operational 2026-10-16 06:36 UTC
Auth
pausedcovered by the control-plane probe; the current result is fetched in your browser · last measured operational 2026-10-16 06:36 UTC
Session control plane
pausedcovered by the control-plane probe; the current result is fetched in your browser · last measured operational 2026-10-16 06:36 UTC
Session fleet
pausedcovered by the control-plane probe; the current result is fetched in your browser · last measured operational 2026-10-16 06:36 UTC
Workspace transfer
not yet probedper-component monitoring planned, covered by the control-plane probe until then · no per-component probe yet; fleet-scoped
Checkpoint store
not yet probedper-component monitoring planned, covered by the control-plane probe until then · no per-component probe yet; fleet-scoped
Gateway and inference
not yet probedper-component monitoring planned, covered by the control-plane probe until then · no per-component probe yet; fleet-scoped
Ledgers
pausedcovered by the control-plane probe; the current result is fetched in your browser · last measured operational 2026-10-16 06:36 UTC
Forms and email
not yet probedper-component monitoring planned, covered by the control-plane probe until then · no live probe; outbound mail is evidence-backed by a delivery gate, which is not an uptime measurement (record )
Status feed
pausedcovered by the control-plane probe; the current result is fetched in your browser · last measured operational 2026-10-16 06:36 UTC
Cruise jobs
pausedcovered by the control-plane probe; the current result is fetched in your browser · last measured operational 2026-10-16 06:36 UTC
Uptime history is unavailable. No day-by-day record is kept, so none is drawn and none is invented.
A release receipt is not a health check: the release version and its date live in the footer's release-details link, and nothing on this page reads from them. Checkpoint durability is a data claim rather than an uptime claim and is gated separately on Proof.
Incident history.
Dated incident summaries state affected customer behavior, known impact, remediation status and remaining limitations. Historical incidents do not by themselves establish current exploitability.
-
Historical usage-settlement incident
On 17 September 2026, monitoring found that metered session usage had not reduced prepaid credit, leaving usage above credited balances. A production correction settled outstanding usage once, recorded the funds restriction, and returned the monitor to clear; the incident was marked resolved that day. No affected session was running, so parking a running session at exhaustion was not demonstrated. A later amendment corrected an overstated monitoring count. Account identities, balances and operational recovery instructions are omitted; the customer-impacting billing failure and test limitation remain visible.
17 September 2026 · Historical record; any production observation is limited to the dated test described below.
record ·
3ddfb9e8316d -
Historical credit-authorization incident
The incident record describes a defect, present from 31 August 2026 and found on 16 September, that allowed a signed-in customer to alter its own credit balance. The 17 September record reports passing laboratory checks for a correction that restricts credit changes while preserving legitimate purchases and ledger reads. It explicitly leaves production deployment verification and ledger review as follow-up work in this version. The record does not establish whether customers misused the defect. Production follow-up belongs to separate later evidence; this historical disclosure is not proof of current exploitability.
31 August 2026 · Historical record; any production observation is limited to the dated test described below.
record ·
5afd378db114 -
Resolved historical status-display incident
On 15 September 2026, KeepState identified a website error that displayed fresh service-health information as paused. The affected site release had gone live on 14 September. The incident record reports that the service itself was operational; the fault was in how the website interpreted freshness. It records a display correction, regression checks and monitoring changes shipped with the incident release. This account distinguishes a misleading status display from a service outage. It is historical incident evidence, not a live availability measurement.
15 September 2026 · Historical evidence; the scope and environment are described in the recorded result.
record ·
f5b4f3651e83
Subscribe to incidents
One email when an incident opens, one when it closes. Nothing else on this list.
Service-incident notices are not an editorial newsletter. Every message carries a one-click unsubscribe, and using it stops delivery immediately.
Machine-readable
Poll it, page on it, put it in your own dashboard. Same fields as the footer system line.
curl keepstate.ai/status.json What counts as an incident
Any degradation a customer could notice, whether or not anyone reported it.