Scheduled Runs

Run agents automatically on a schedule. Perfect for daily reports, monitoring dashboards, data sync jobs, and any recurring task you want to automate without writing cron infrastructure.

How It Works

1. Open your agent → Settings tab → Scheduled Runs panel
2. Enable the schedule and choose a frequency
3. Write a prompt that the agent receives as the user message each run
4. Pick a timezone
5. Save — the agent will start running on the next due time

Each scheduled run is billed the same as a normal agent execution and appears in your Run History with full step-by-step observability.

Schedule Types

Configuration Schema

Schedules are stored as part of the agent document. You can also manage them via the API:

// Update schedule via Cloud Function
const { httpsCallable } = require('firebase/functions');

const updateSchedule = httpsCallable(functions, 'updateAgentSchedule');

await updateSchedule({
  agentId: 'my-daily-reporter',
  schedule: {
    type: 'daily',
    enabled: true,
    input: 'Generate today\'s analytics summary report',
    hour: 9,
    minute: 0,
    timezone: 'America/New_York'
  }
});

Schedule Prompt

The input field is sent as the user message to your agent on each scheduled run. Write it as if you're typing into the test panel:

"Generate a summary of all customer support tickets from the last 24 hours. Include:
- Total tickets opened and closed
- Average resolution time
- Top 3 most common issues
- Any urgent items still unresolved"

Error Handling

• Automatic retry: If a scheduled run fails, the failure is recorded and the schedule continues on the next interval.
• Auto-disable: After 5 consecutive failures, the schedule is automatically disabled to prevent wasted credits. A warning badge appears in the agent editor.
• Re-enable: Fix the issue (e.g. add credits, fix the prompt), then toggle the schedule back on in Settings.

Observability

Scheduled runs appear in your agent's Run History just like manual or API-triggered runs. Each run includes:

• Step-by-step execution trace
• Token usage and cost breakdown
• Model used per step
• Tool invocations and results
• Latency metrics
• Export to JSON/CSV

The schedule panel also shows the last run status and any error messages at a glance.

Billing

Scheduled runs are billed identically to regular agent executions. Each run deducts credits based on actual token usage. If your account runs out of credits, the scheduled run will fail with an insufficient_credits error and count toward the consecutive failure limit.

Tip: Enable Auto-Refill in Billing settings to ensure scheduled agents never run out of credits.