Cron Jobs

Scheduled jobs run inside --serve mode alongside the Discord bot and HTTP API.

Configuration

Add jobs under cron.jobs in config.yaml:

cron:
  enabled: true
  jobs:
    - name: "daily-email-summary"
      schedule: "0 9 * * *"          # standard cron syntax
      prompt: "Summarize my unread emails from the last 24 hours"
      profile: "email-checker"
      delivery:
        channel: "log"               # or "discord" / "discord-dm"

Set cron.enabled: true and run with --serve — the scheduler starts automatically.

Job modes

Wake agent (default)

wakeAgent: true runs the full agent loop with the configured prompt, then delivers the response via the configured delivery channel.

Add note

wakeAgent: false injects the prompt as a user message into the session without running the agent loop. Useful for queuing context for the next interaction.

Delivery channels

• log (default) — prints the response to stdout
• discord — sends to a Discord channel (requires delivery.target channel ID)
• discord-dm — sends as a DM to the bot owner

Job options

| Field | Description | |-------|-------------| | name | Unique job identifier | | schedule | Cron expression (standard 5-field syntax) | | prompt | The message sent to the agent | | profile | Named profile to use (optional) | | wakeAgent | Run agent loop (default true) or just inject message | | newSession | Start a fresh session each run (default false) | | delivery.channel | Output channel: log, discord, or discord-dm | | delivery.target | Discord channel ID (required for discord delivery) | | hooks | Job-specific hooks (appended to profile hooks) |

Prompt templating

Cron prompts support template variables:

| Variable | Description | |----------|-------------| | {{last_run}} | ISO timestamp of the last run | | {{last_run_epoch}} | Unix timestamp of the last run | | {{last_response}} | The agent's response from the last run | | {{next_task}} | First "Next:" line from the profile's goals.md |

Example with hooks

cron:
  enabled: true
  jobs:
    - name: "morning-briefing"
      schedule: "0 8 * * *"
      prompt: "Give me a morning briefing based on the context below."
      profile: "researcher"
      hooks:
        beforeRun:
          - tool: gmail
            args: { action: "check", query: "newer_than:1d" }
            skipIf: "no new messages"
          - tool: web_search
            args: { query: "AI news today" }

Job state

Job state (last run time, last response) is tracked in the cron_jobs database table and persists across restarts.