Skip to content

Status

One connector is degraded.

Driven by real health probes, not a switch somebody remembers to flip. Per-connector status is published separately, because “Zendesk sync is degraded” is a very different message from “Kestrel is down”.

99.993% mean availability across all components over 90 days

Platform

  • API

    api.kestrel.io

    operational

  • Eyrie console

    eyrie.kestrel.io

    operational

  • Stoop orchestration

    Diagnosis and action selection

    operational

  • Creance

    Policy evaluation

    operational

  • Flightlog ingestion

    Audit write path

    operational

Agent

  • Agent check-in ingestion

    ingest.kestrel.io — Bells telemetry and action dispatch

    operational

  • Recall stand-down

    standdown.kestrel.io — independent of orchestration, by design

    operational

  • Mews update service

    dl.kestrel.io — agent and bundle distribution

    operational

Connectors

  • Jess — Zendesk

    Webhooks healthy, last delivery 14s ago

    operational

  • Jess — ServiceNow

    Elevated write rejections on one region since 11:40Z. Polling unaffected; resolutions are landing up to 60s late.

    degraded

  • Jess — Freshservice

    Webhooks healthy

    operational

  • Jess — Jira Service Management

    Webhooks healthy

    operational

  • Jess — ConnectWise

    Polling — this connector has no webhook support

    operational

Incident history

  • 2026-08-01

    Elevated ServiceNow write rejections in eu-west

    Investigating

    From 11:40Z a subset of ServiceNow writeback calls in eu-west are being rejected by an upstream business rule. Inbound polling and diagnosis are unaffected; resolution notes are queued and will be replayed. No customer action needed.

  • 2026-06-01

    Stoop orchestration degraded for 41 minutes

    Resolved · postmortem published

    A schema migration on the workflow store caused diagnosis to queue rather than fail. No incorrect actions were executed, and Creance evaluated normally throughout. Agent check-in and Recall were unaffected — Recall runs on separate infrastructure precisely so that an orchestration incident cannot take the stop button with it.

NOTE

In production this page runs on infrastructure fully independent of the platform, with separate DNS — a status page that goes down with the product is worse than no status page. In this build it is served from the same host as everything else, so it does not yet meet that bar. We would rather say so here than let you assume otherwise.