Schedules

Run a task on a clock instead of from your code: every weekday morning, every 15 minutes, or once next Friday. Add a schedule to a task and Blazing Agents starts each task run for you.

Create a scheduled task

This task runs at 9:00 London time every weekday. Set AGENT_ID to an agent with a provider and model, such as the one from the quickstart.

import { BlazingAgents } from "@blazingagents/sdk";

const client = new BlazingAgents({
  apiKey: process.env.BLAZING_AGENTS_API_KEY!,
});

const { task } = await client.tasks.create({
  agentId: process.env.AGENT_ID!,
  name: "Morning summary",
  prompt: "Summarize yesterday's support tickets.",
  schedule: {
    kind: "cron",
    config: { expression: "0 9 * * 1-5", timezone: "Europe/London" },
  },
});
console.log(task.id, task.schedule);

You see the task ID and the saved schedule. At each scheduled time a new run appears in client.tasks.listRuns({ taskId }) (client.tasks.list_runs(task_id) in Python). Read its result the same way as any task run.

Choose a schedule

For example, { kind: "interval", config: { everyMs: 15 * 60_000 } } runs every 15 minutes. In Python, use every_ms and stagger_ms.

staggerMs delays each cron run by up to that many milliseconds. Each task gets its own fixed delay within that limit, so many tasks that share a cron time do not all start at once.

Change or pause a schedule

• Replace the schedule with client.tasks.update({ taskId, schedule }), or pass schedule: null to make the task on-demand only.
• Pass enabled: false to pause scheduled runs and keep the schedule. Pass enabled: true to resume it.

After a change, runs that were due under the old schedule no longer start.

If you disable the task's agent, scheduled times pass without creating runs. Recurring schedules pick up at their next time once you enable the agent again. A one-time schedule whose time passed while the agent was disabled does not run later.

What to expect

• One run at a time. If a run is still active when the next time arrives, that time is skipped.
• No catch-up. Missed cron and interval times are skipped, not run in a burst later. Intervals continue at their next regular time.
• Late one-time runs. A one-time schedule that was missed because of a platform delay may start as soon as the platform recovers.
• Blocked runs. If your quota is exhausted, a scheduled run ends as blocked and the next scheduled time tries again. See usage and quotas.

Next

• Task runs to read results and cancel runs.
• tasks.update() for every schedule field.