Skip to content

Legal

Service Level Agreement

Monthly uptime, exclusions, service credits by tier, support response clocks, and claim rules for nox.markets. Starter and Growth have no availability SLA.

Last updated: 16 September 2026
Provider: NOX AETERNA GROUP LLC, a Wyoming limited liability company, trading as nox.markets, 5830 E 2nd St, Ste 7000 #34506, Casper, WY 82609, USA
Governing law: State of Wyoming, USA
Related: /status, /trust, Terms of Service

This Service Level Agreement (SLA) is part of the Agreement for paid use of the hosted nox.markets application and production API. It does not apply to the marketing website except where that site is required to reach /status.

Sole remedy. Service credits described here are Customer’s sole and exclusive remedy for availability (and, where stated, for missed support response). Credits apply to the next invoice; they are not wired. Cap: 50% of that month’s platform subscription (not credit overage, not one-time licenses).


1. Who has an availability SLA

tierAvailability SLARPORTO
starterNone. Best effort. $0 availability credits. Support response SLA only24 hours (last daily snapshot + WAL best-effort)8 hours, business hours
growthNone. Best effort. $0 availability credits. Support response SLA only24 hours8 hours, business hours
scale99.9% monthly1 hour4 hours
enterprise99.95% monthly, with credits15 minutes1 hour

A 30-day month is 720 hours = 43,200 minutes.

  • 99.9% allows 43.2 minutes unplanned downtime in that month before the SLA is missed.
  • 99.95% allows 21.6 minutes.

Self-hosted products are excluded. Uptime is Customer’s. License telemetry availability is best effort.

Starter and Growth paid promises for this document are support response times, not uptime.


2. Uptime definition and measurement

Monthly uptime U = (total minutes in the month − excluded minutes − Downtime minutes) / (total minutes − excluded minutes).

Downtime means the production control plane (authentication, workspace API, and the hosted app required to start or monitor runs) is unavailable to a material portion of customers in the workspace’s region, as measured by our status provider.

A single product slug failing while the control plane is up is a P2 product incident, not automatic Downtime for the platform SLA, unless the failure is caused by our control plane.


3. Exclusions (not Downtime)

The following do not count as Downtime and are excluded from the denominator or numerator as applicable:

  1. Scheduled maintenance in a published window: weekly window of up to 2 hours, announced 72 hours ahead, typically unused. Maintenance is planned or it counts. If we needed a 4-hour window and called it unplanned, Enterprise would burn its monthly budget twice over (240 minutes vs 21.6).
  2. Customer-side IdP, OAuth, or connected-SaaS failures (QuickBooks, HubSpot, mail providers, Customer’s DNS).
  3. Model-subprocessor outages we did not cause. We pass through their status. We credit only if our control plane was down. If OpenAI/Anthropic/Google is down and we are up, orchestration may fail for model-calling products; that is not our availability credit unless we are down.
  4. self-hosted deployments.
  5. Customer configuration, exhausted spend ceilings, circuit breaker, AUP suspension, unpaid invoices, or paused products.
  6. Force majeure as defined in the MSA / Terms.
  7. Beta, trial (except that trial still receives Growth support clocks, not Scale uptime credits), and features marked experimental on the product page.
  8. Internet, Customer VPN, or browser issues outside our edge.

Backups: continuous WAL for the primary database; encrypted snapshots every 24 hours; 35-day retention; stored in Germany; no backup copies outside Germany. Restore tested quarterly; first test Planned 2026-12-31.

Disaster recovery at launch: the Service runs on a single Hetzner server in Nuremberg, Germany.

This SLA does not claim multi-region active-active.


4. Service credits (Scale and Enterprise only)

Measured monthly on the production API and app.

Monthly uptime UScale creditEnterprise credit
U ≥ SLA00
SLA > U ≥ 99.0%10% of that month’s platform subscription10%
99.0% > U ≥ 98.0%25%25%
U < 98.0%50%50%

Credits do not apply to credit overage, one-time licenses, self-hosted license fees, professional services, or VPC add-ons unless an order form says otherwise.


5. Claim procedure

Customer must open a support ticket within 30 days of month end identifying the incident windows and workspace. We will compare against /status and internal monitors. Approved credits appear on the next invoice. Failure to claim in time waives the credit for that month.


6. Support response SLA

Business hours: 09:00–18:00 US Eastern, Monday–Friday, excluding US federal holidays.

tierChannelFirst responseP1
starterEmail24 business hoursSame (no 24/7)
growthEmail + chat8 business hoursSame; 60-minute onboarding call is scheduling, not Sev-1
scaleNamed Solutions Architect, Slack Connect4 hours1 hour P1
enterpriseDedicated CSM + Solutions Architect4 hours standard1 hour P1, 24/7

P1 = production down or suspected customer-data exposure.
P2 = a single slug failing, workaround exists.
P3 = how-to.

Personal-data breach notification to the workspace owner is a DPA clock (≤ 72 hours after we confirm), not this support table. P1 availability customer notice: 24 hours from declare, and /status. P2 security without personal data: 72 hours. See /security.

Missed support response:


7. Relationship to other documents

SOC 2, ISO 27001, and HIPAA are not availability claims and are not Held as of 15 September 2026. See /trust.

This SLA does not warrant that model output is correct. Wrong postings to Customer’s books are not Downtime.