Skip to main content

Automations

Start Agent work on a schedule and verify each result in Run history.

Use an automation for repeatable work that should start without a manual prompt, such as periodic checks, code maintenance, tests, or report generation. Qoder starts an Agent task on schedule and collects each result in Run history.

Decide whether to automate the task

The task should be able to finish without live clarification. It needs clear inputs, an observable completion condition, and resources accessible from its execution environment. Local execution also needs a stable working directory; cloud execution cannot depend on your local environment. If each run requires a new judgment or sensitive confirmation, test it as a regular task first.

Open Automations

Select Automations in the sidebar. Choose an entry point based on how much of the workflow is already defined:
  • Select Create with Qoder when you can describe the outcome but want Qoder to help structure the automation.
  • Select New automation when the schedule and task are already clear and you want to fill out the form directly.

Choose where to run

When creating an automation, choose local or cloud execution:
  • Run locally runs on your computer and suits tasks that need local working directories, files, or Skills.
  • Run in cloud runs in a cloud environment. You can select existing cloud Skills and connectors available to your account and add local files or images as task attachments when cloud uploads are available. Attachments are uploaded when you save the automation; local folders and your full working environment are not uploaded. Keep the goal, inputs, and completion criteria in the task instructions.
Cloud execution is supported on individual plans (Free, Lite, Pro, Pro+, and Ultra) and business plans (Teams and Enterprise).

Models for cloud automations

When editing a cloud automation, choose a model from the cloud task catalog. The list depends on your account, organization, and cloud capabilities, and can include enterprise models you have access to. Configuring a custom API-key model or external provider on your computer does not add it to the cloud catalog.

Create an automation

1

Set the name and schedule

Enter an Automation name, choose local or cloud execution, then select the Schedule, time, and Time zone. For local or cloud recurring schedules, enable Set expiration date to stop after a defined period; after that date, it pauses and no longer creates executions. The expiration date cannot precede the next valid execution. It applies to recurring schedules, not one-time runs. Check the time zone and next execution before saving.
2

Describe the work

In Agent prompt, state the outcome, scope, constraints, and completion check. For local execution, add context or files and select the work folder and model when needed. For cloud execution, enter the task instructions, select the existing cloud Skills and connectors needed by the run, and add any required files or images through the attachment picker. An automation can retain up to 20 attachments totaling 100 MiB, including attachments already saved with it. Remove attachments you no longer need before saving. Prepare missing cloud resources in Settings → Cloud environment.
3

Choose the task output

Select Independent task when every execution should create a separate task. Select Merged task when results should accumulate in one task and earlier history should be included in the next execution.
4

Review execution authorization

For local execution, review Execution authorization, which enables unattended operation. When it is enabled, a scheduled Agent can call tools without waiting for the confirmations used in a regular task. Limit access to the workspace and operations required for the automation. This switch is not shown for cloud execution.
5

Create the automation

Review the execution location, schedule, and instructions. For local execution, also check the working directory and authorization scope. Then select Create.
Write the instructions as a handoff another person could execute directly. Include the input source, allowed changes, prohibited actions, validation method, and expected output.
After you save an automation through a conversation, its result card appears in the reply. Check the name, execution location, and schedule, then select the card to open the automation details. View a local automation on the device that saved it. For a cloud automation, use the account and organization with access. Open Run history to check what an individual run produced.

Create an automation on a connected device

Enter / in a task conversation on a connected device, choose the automation entry, and describe the schedule, working directory, task, and completion criteria. Check the conversation’s execution device and access scope before submitting the request. The automation and its run history are stored on the execution device. Manage the schedule and review results there. This entry works in local and connected Qoder-device conversations, not SSH or WSL conversations.

Keep local schedules runnable

Qoder must be running at the scheduled time and able to reach the required network, accounts, and files. Enable Keep device awake to prevent system sleep while Qoder is running and reduce missed schedules or interrupted Agent sessions.
Keeping the device awake does not replace reliable power, network access, or account availability.
Where capabilities and enterprise access allow, local automations can use built-in collaboration tools with the current signed-in account without separately installing or signing in to a standalone CLI. Existing approval and business-authorization rules still apply.

Verify the execution

Do not treat a created automation as proof that a scheduled run succeeded. Return to Automations and open Run history:
  • Use Group by day, Group by week, or Group by month to review different periods.
  • Use Filter by automation when the list contains several automations.
  • Use Filter by status to focus on failed or pending executions.
  • Select Refresh when the latest state is not visible.
Open an execution to inspect the Agent output and validation result. For automations that change code or external data, require reviewable changes and verification evidence in the task instructions. Run history loads in pages. After archiving, refresh lists in other windows if they still show the old state.

Maintain automations

Edit the automation when its instructions or working directory change. To interrupt a run or retire a schedule, choose the operation that matches your goal:
GoalAction
End the current runOpen its conversation, stop it, and wait for status feedback.
Prevent future scheduled runsPause the automation.
Remove the definitionDelete the automation after reading the confirmation.
Stopping a run does not automatically pause its schedule. Cloud stops require remote confirmation; recheck status when the outcome is uncertain. You can still inspect existing output. Before continuing or rerunning, check for duplicate external operations.