Automations

Automations let you define tasks that run automatically — on a schedule, in response to GitHub events, or on demand.

Trigger Types

Manual

Run the automation whenever you want by clicking Run Now. Useful for tasks you trigger occasionally, like code audits or dependency updates.

Cron

Schedule automations using cron expressions. Shellarium shows a live countdown timer to the next run.

Examples:

• Every hour — 0 * * * *
• Weekdays at 9am — 0 9 * * 1-5
• Every Monday — 0 0 * * 1

GitHub Events

Trigger automations in response to GitHub activity:

• New pull request opened
• Issue created or updated
• Push to a specific branch

You can scope triggers to specific repositories and branches.

Creating Automations

From Scratch

1. Click the + button in the Automations sidebar section
2. Give it a name and describe the task
3. Choose a trigger type and configure it
4. Select the AI provider and model
5. Save — the automation is ready

From an Existing Chat

After a successful agent session, click Save as Automation to turn that chat into a reusable automation. The task description and configuration are pre-filled.

AI Configurator

Each automation includes an AI chat assistant that helps you set up the configuration interactively. Ask it to adjust the prompt, change the schedule, or tweak provider settings — it updates the automation config in real time.

Configuration Options

• Provider & model — choose which AI agent and model to use
• Thinking & reasoning — enable extended thinking for complex tasks
• MCP servers — attach MCP servers for additional tool access
• MCP auth policy — choose how unattended runs handle unauthorized MCP servers: ignore (continue) or fail (stop run)
• Orchestrator mode — enable sub-agent delegation
• Project scope — target specific projects and repositories
• Concurrent runs — limit how many instances can run in parallel
• Retries — configure retry behavior for failed trigger processing
• Max concurrent & cooldown — control overlap and pacing for bursty trigger streams

Monitoring Runs

Each automation tracks its run history in the Runs tab:

• Run status (running, completed, error)
• Duration and cost
• Link to the workspace with full agent output and diff

Click any run to see the full workspace detail with output, diff, and merge options.

Event Queue

The Event Queue tab shows pending and processed automation events. When a GitHub event or cron trigger fires, the event is added to a persistent dispatch queue. Events are processed reliably — even if the app restarts, queued events are preserved and dispatched on next launch.

Unhandled events are marked with a badge so you can quickly see what needs attention.

The queue also works with retry scheduling and automation concurrency limits, so events can be deferred instead of dropped when a run slot is unavailable.

Hooks

Automations support a universal hooks system. You can attach custom hooks to the automation lifecycle — for example, running a setup script before the agent starts or a cleanup step after completion.

Sidebar

All automations appear in the global Automations section in the sidebar, regardless of which project is currently open. Each automation shows its trigger type icon and last run status.