Automations

Automations let you run AI-powered tasks within your workspace. Define a prompt, pick a trigger — a recurring schedule or a manual Run now button — and EM Kit executes it using your workspace data and connected integrations to produce reports, summaries, or alerts.

Requirements

Automations use the same AI provider as EM Copilot. Before creating an automation:

1. Install Claude Code or Codex CLI and authenticate in your terminal.
2. Configure the provider in Settings → EM Copilot.

Tabs

The Automations page has three tabs:

• Automations — List of all automations in the workspace.
• History — Run history showing past executions and their results.
• Templates — Pre-built automation templates to get started quickly.

Creating an automation

1. Click Create automation (or pick a template from the Templates tab and click Use template).
2. Fill in the form:
   • Name — A short label for the automation.
   • Prompt — The instruction the AI will execute. Reference your workspace data, integrations, or anything the EM Kit MCP server exposes.
   • Trigger — Choose Schedule for recurring runs or Manual for on-demand only. See below.
   • Schedule mode (Schedule trigger only) — Choose Weekly or Interval.
   • Enabled (Schedule trigger only) — Toggle the automation on or off.
3. Click Create.

Schedule trigger

Schedule-triggered automations run on a recurring cadence while the EM Kit desktop app is open.

Weekly schedule

Pick one or more days of the week and a time. The automation runs at the chosen time on each selected day, in your local timezone.

Interval schedule

Set a number of hours (1–168). The automation runs repeatedly at that interval.

Manual trigger

Manual automations have no schedule and never run on their own. They sit idle until you click Run now — either from the row in the Workspace Automations tab or from inside the edit modal. Use this when:

• The work is bursty or context-dependent (e.g. “draft a recap for the project we just shipped”).
• You want a saved prompt + agent + model combo you can re-fire on demand without retyping it.
• The cadence isn’t predictable enough to commit to a schedule.

Manual automations are always considered enabled; the Enabled toggle is hidden because there is no scheduler to gate.

Agent, model, and reasoning

Expand Agent, model and effort in the create/edit modal to control how the automation runs:

• Agent — Pick the agent whose persona and skills the automation should use. The agent is required; the workspace default is selected by default. Switching agents changes both the tone (system prompt) and the available skills for every scheduled run.
• Model — Optional. Choose a different AI model than the workspace default.
• Reasoning effort — Optional. Set to Low, Medium, High, or Max to control how much the AI “thinks” before responding.

Model and reasoning are optional — automations inherit the workspace defaults when not overridden.

Running an automation manually

Every automation — whether its trigger is Schedule or Manual — can be fired on demand:

• From the list — Click the Run now button on the right side of the automation’s row in the Workspace Automations tab.
• From the edit modal — Open the automation and click Run now at the bottom of the modal.

Manually triggered runs are tagged as manual in the history (versus schedule for runs the scheduler started).

Enabling and disabling

The Enabled toggle applies to Schedule-triggered automations only. Toggle it off to pause an automation without deleting it; the configuration and history are preserved but the scheduler will skip it. Manual automations don’t have an Enabled toggle — they only run when you click Run now, so there’s nothing to pause.

History

The History tab shows the last 200 runs across all automations:

• Automation name — Which automation ran.
• Date — When the run started.
• Status — Running, Success, or Failure.

Click a run to inspect its full output (rendered as markdown), any errors, and the AI’s extended thinking.

You can delete individual history entries using the delete icon.

Templates

The Templates tab provides ready-to-use automation configurations:

Template	Schedule	What it does
Daily bot	Weekdays at 09:00	Summarizes team achievements
Weekly achievements recap	Friday at 16:00	End-of-week summary
Missed worklog scan	Friday at 17:00	Identifies missing work log entries
1:1 Prep Brief	Weekdays at 08:00	Prepares context for upcoming 1:1s
Career Growth Check-in	Monday at 09:00	Monthly skill gap analysis
Quiet Contributor Spotlight	Friday at 11:00	Surfaces unrecognized contributions
PR Review Bottleneck Detector	Weekdays at 10:00	Finds stale pull requests
Weekend/After-Hours Work Alert	Monday at 08:00	Detects burnout patterns

Click Use template to pre-fill the creation form with the template’s name, prompt, and schedule. You can customize everything before saving.

Example prompts

• “Summarize the team’s work log entries from this week and highlight any blockers.”
• “Check if any team member has not logged work in the past 3 days.”
• “List open issues assigned to my team that are past their due date.”
• “Prepare a brief for my 1:1 with each direct report, including recent feedback and objectives progress.”

Troubleshooting

Automation shows “Paused” in the next run column

The automation has a Schedule trigger but is disabled. Open it and toggle Enabled on. (If the column instead shows —, the trigger is Manual by design — there’s no next run because it only runs on demand.)

Automation fails with an error

1. Open the run in the History tab to see the error message.
2. Make sure the AI provider is still configured and authenticated — run which claude or which codex in your terminal.
3. Run emkit doctor to verify the CLI bridge is healthy.

Automations surface the same structured error messages as the Copilot drawer. The most common ones:

Message	What it means	What to do
”…did not respond within 600s”	Wall-clock timeout hit.	Shorten the prompt, lower reasoning effort, or check that the provider CLI isn’t waiting for login. See EM Copilot troubleshooting for the full checklist.
”…is not authenticated”	The provider CLI lost its session.	Run claude login or codex login in your terminal.
”…stopped at the max-turns limit”	The AI used its full tool-call budget without producing a final answer.	Simplify the prompt or raise the limit in your Claude Code config.
”…returned an empty response”	The CLI exited cleanly but produced no text.	Re-run the automation; if it recurs, check provider status in Settings → EM Copilot.

No runs appear in history

1. For Schedule triggers, confirm the automation is enabled and has a valid schedule.
2. For Manual triggers, history only fills in once you click Run now at least once.
3. Make sure the EM Kit desktop app is running — the scheduler only runs while the app is open.