Automations

Automations let you run agent prompts on a schedule — useful for recurring tasks like code reviews, dependency checks, documentation updates, or any routine work you want an AI agent to handle automatically.

How it works

Each automation is tied to a workspace and defines:

Name — a label for the automation.
Agent type — which agent to use (Cursor, Claude, or Mistral Vibe).
Prompt — the instruction the agent receives on each run.
Interval — how often to run, in minutes. Common presets include 15 minutes, 1 hour, and daily.
Enabled — toggle the automation on or off without deleting it.

The server maintains a scheduler that tracks next run and last run times for each automation.

Manual triggers

You can trigger any automation immediately outside its regular schedule — useful for testing a new automation or running one on demand.

Run history

Every automation run is recorded with:

Status — whether the run succeeded or failed.
Agent response — the full output from the agent.
Changed files — which files the agent modified during the run.
Errors — any issues that occurred.

The UI polls for updates while a run is in progress, so you can watch results come in.

Use cases

Daily code review — “Review recent commits for potential issues.”
Dependency updates — “Check for outdated npm packages and suggest updates.”
Documentation sync — “Update the API docs to match the current route definitions.”
Health checks — “Verify all tests pass and report any failures.”
Code cleanup — “Find and fix lint warnings in the src/ directory.”

Requirements

The Nova Code server must be running for scheduled automations to fire. If the server is stopped, missed runs are not retroactively executed.
The configured agent must be authenticated and available (see Settings).

Sessions — automations use the same agent infrastructure as sessions
Workspaces — each automation belongs to a workspace
Installation — keep the server running for schedules