Automation Basics
What is an automation?
An automation is a saved, reusable multi-step workflow. Once saved, it runs autonomously — triggered by a schedule, a webhook, an API call, or a manual run. The automation executes its steps in order, passing data between them, without any user interaction.
Anatomy of an automation
- A title and description
- A sequence of typed steps (the workflow logic)
- An optional success_criteria string describing what a successful run looks like
- Optional scheduling (cron expression, interval, one-time datetime, or webhook URL)
- Error handling settings controlling how the platform responds to failures
Steps and data flow
Steps are the building blocks of an automation. Each step has a type (call_api_endpoint, run_sandboxed_code, use_mcp_tool, if, switch, loop, merge, wait, generate_pdf, generate_word, combine, seek_human_approval_for_next_step, trigger_automation), a description, and type-specific parameters. The output of each step is automatically available to subsequent steps via the dataPrev variable.
Agents vs. automations
An AI agent responds to user input in real-time conversation. An automation executes a pre-defined workflow autonomously on a trigger. Use an automation when you need a workflow that runs on a schedule or in response to an event without any user prompt. Use an agent when you need a workflow that responds to natural language from a user.
Active automations limit
Depending on your plan, there may be a limit on how many automations can be active simultaneously. The Starter plan allows 1 active automation. Paid plans increase this limit. You can create more automations than your active limit — simply disable unused ones to stay within the limit.