Skip to main content
The Schedules section (/platform/schedules) is the operator surface for every scheduled job the platform runs: the recurring platform crons (workflow watchdog, retention sweepers, budget resets, drift alerts, model catalog sync, governance reminders, sandbox heartbeats, Judge metrics aggregation and others), the workflow triggers across the org unit, and pending one-off jobs. This entire page is admin-only. All three tabs — Recurring, Upcoming, and the operator-wide Triggers rollup — require Platform admin; non-admins are redirected to the builder scheduling surface at Production › Workflows › Schedules, where builders manage their own workflow schedules per-workflow. The Triggers tab here is the platform-wide trigger rollup, not the builder’s scheduling home.

What you can do

  • Check the health of every platform cron on the Recurring tab, failing and overdue ones first.
  • Open a cron’s run history, with each run’s outcome, duration and error.
  • Pause, resume or run a cron now, with platform:manage.
  • See what fires next: up to 72 hours of scheduled work on the Upcoming tab.
  • Review every workflow trigger in the org unit on the Triggers tab.

What you can see

The page has three tabs. The active tab is persisted in the URL (?tab=recurring / ?tab=upcoming / ?tab=triggers) so links are shareable.
A health-first inventory of every platform cron Forge ships. A KPI health band sits above the table, and rows are ranked by attention — anything failing or overdue floats to the top.
Schedules — Recurring tabSchedules — Recurring tab

Schedules → Recurring — Forge Console at /platform/schedules?tab=recurring

The KPI health band summarizes the fleet at a glance: healthy, needs-attention (failing + overdue), degraded, paused, and due-in-the-next-hour counts.Filters above the table narrow by category, health, and a case-insensitive name search; they compose. Each row exposes Pause / Resume and Run now controls, and clicking a row opens the drill-in drawer (see below).

Drill-in drawer

Clicking a recurring row opens a side drawer with the cron’s run history: each run’s trigger (cron tick or operator manual), start time, duration, outcome, and — on failure — the captured error text. A duration trend surfaces slow-downs over time. The drawer also carries the Pause / Resume and Run now controls.

How the data flows

Every platform cron runs through one dispatcher. It checks whether the cron is paused (a cron that has never been paused or resumed counts as enabled) and records the run before the job starts. A follow-up job then records each run’s outcome, duration and error. Run history is kept for 30 days.
  • Recurring shows one row per platform cron: its next fire time, worked out from the cadence; whether it is enabled; its health; and its last 20 run outcomes, read from that run history.
  • Upcoming merges the scheduled work from its three sources in fire-time order, up to 72 hours ahead. A paused cron shows no upcoming fires.
  • The drawer reads one cron’s run history, and the row controls pause, resume or run a cron now.
Reading the roster, the upcoming list and a cron’s history needs platform:read. Pausing, resuming and running a cron now need platform:manage. The Console page is also limited to platform administrators: anyone else is redirected away from it. The permissions apply to every request, including requests made outside this page. Relative times on the page refresh together every 30 seconds.

Scheduling your own work

Platform crons ship with Forge. You cannot add one from this page, the Console or the API. To run your own work on a schedule, give a workflow a recurring (cron) or one-shot schedule from its Schedule tab: see Scheduling workflows. Integrators can create the same schedules through the API with createTrigger.

What’s not on this page

  • Editing a platform cron’s cadence. Cadence ships with Forge and changes only with a release. Operators can pause / resume and run-now, but not re-time a cron from the console. (Workflow-trigger CRUD, including cadence edits, lives on the Triggers tab.)
  • Convex internal scheduled work older than the next 72 hours. The pending-jobs portion of the Upcoming tab is bounded by the window selector for performance.

Key state to watch

  • Needs attention in the KPI health band — failing and overdue crons, which also rank first in the Recurring table.
  • Failures column — a red non-zero count means failures among the last 20 runs; open the row’s drawer for the error text.
  • Paused crons — a paused cron shows — for its next fire and projects nothing on the Upcoming tab.
  • Paused triggers on the Triggers tab — a trigger whose creator lost workflows:manage is paused with a workflow_trigger_refused audit event.
  • Workflow schedules for operator-created workflow triggers and their cadence.
  • Runs for executions that a schedule has already launched.