Skip to main content
Scheduled Jobs run Dexter prompts on a recurring or one-off schedule, creating a new conversation each time they fire. This lets you automate recurring factory checks, report generation, simulation runs, and pipeline refreshes without manual intervention.

​
Creating a Scheduled Job

Tell Dexter what you want automated and when. Dexter translates natural language into a schedule, confirms the details with you, then writes the job:
  • “Every weekday at 5am, run the simulation and email me a PDF of the results”
  • “Every Monday morning, refresh the production data pipeline”
  • “Run a Monte Carlo batch on the first of every month”
  • “In 2 hours, generate a quarterly throughput report”
Dexter handles the scheduling syntax — describe the cadence in plain language and Dexter interprets it.
You sayDexter interprets
”Every weekday at 7am”Monday through Friday, 7:00 AM
”Every Monday morning”Monday, 9:00 AM (default morning time)
“The first of every month at midnight”1st of each month, 12:00 AM
”Every 6 hours”Every 6 hours from creation time
”In 2 hours”One-off job, fires once at the scheduled time
All schedules use your account’s timezone (IANA format, e.g., America/New_York). Dexter confirms the interpreted schedule before creating the job, so you can correct any misinterpretation.
Multi-time schedules like “twice a day at 6am and 6pm” require two separate jobs — a single schedule can’t represent two arbitrary fixed times.

​
Recurring vs. One-Off Jobs

Scheduled Jobs support two modes:
  • Recurring — runs on a cron schedule (daily, weekly, monthly, custom intervals). Keeps firing until you delete it.
  • One-off — runs once at a specific future time. Useful for deferred tasks: “On April 1st at 8am, run the quarterly analysis.”
For complex tasks, create a custom workflow first, then reference it from the scheduled job. This keeps the job prompt clean and the procedure reusable: “Run the weekly quality audit workflow every Monday at 6am.”

​
How Scheduled Jobs Run

When a scheduled job fires:
  1. Dexter opens a new conversation with the job’s prompt as the first message
  2. Dexter executes the instructions autonomously — running simulations, generating reports, refreshing pipelines, or any other capability
  3. Any outputs (reports, exports, data files) are saved as if you’d generated them yourself
  4. The resulting conversation is saved to your conversation history with a status indicator
Scheduled runs appear alongside your interactive conversations, so you can review what Dexter did, examine tool calls, and follow up if anything needs attention.
Scheduled jobs have full access to everything Dexter can do interactively. Each job runs against the factory that was active when you created it — the job is permanently bound to that factory.

​
Managing Scheduled Jobs

Open the Scheduled Jobs section in Dexter’s left sidebar to see all your jobs. Each entry shows the job name, schedule, status, and last run time. From this panel you can edit a job’s schedule or prompt, or delete it permanently.
Pausing and resuming jobs is not currently supported. To temporarily stop a job, delete it and recreate it when you’re ready to resume.

​
Reviewing Past Runs

Every scheduled job execution creates a conversation in your history. To review what a job did:
  1. Open your conversation history in Dexter’s left sidebar
  2. Look for conversations created by the job (they’ll show the job name as the conversation title)
  3. Open the conversation to see the full execution: tool calls, results, any outputs generated
This makes scheduled work fully auditable — you can trace any report or result back through the exact steps Dexter took to produce it.

​
Notifications

Toggle on desktop notifications in Dexter Settings to receive browser notifications when a scheduled job completes or encounters an issue. For delivery to external channels (email, Slack), include the delivery step in the job prompt itself — e.g., “Run the simulation every Monday at 7am and email the PDF report to ops-team@company.com.” External delivery depends on active integrations.

​
Limitations

  • One factory per job. To automate work across multiple factories, create separate jobs in each factory.
  • No interactive input. Scheduled runs are fully autonomous — write prompts that include all the context Dexter needs.
  • No pause/resume. Delete and recreate to temporarily stop a job.

​
Best Practices

  • Keep prompts specific. “Generate a PDF report on last week’s throughput for the CNC line, including utilization and cycle time KPIs” is better than “send me an update.”
  • Build workflows for complex jobs. Create a custom workflow that defines the procedure, then schedule a job that references the workflow by name.
  • Include context Dexter might need. Name specific models, data sources, and naming conventions in the prompt.
  • Review early runs. Check the first one or two runs to confirm Dexter is doing what you expect. Tune the prompt if anything’s off.
  • Use the persistent knowledge base. The more Dexter knows about your operation, the better scheduled runs will match your expectations.
Scheduled jobs execute with your permissions. Any action Dexter takes during a scheduled run has the same effect as if you’d asked for it interactively. Changes to models, BOMs, or configurations take effect immediately.