Step Types

Read Data

Fetch records from any entity in your project.

Configuration

Accessing results

Read Data returns an array. Access individual fields like this:

{{ $anythink.steps.get_users.data[0].name }}
{{ $anythink.steps.get_users.data[0].email }}

You can also pass the full array to a downstream Create Data step for bulk operations.

Create Data

Insert one or more records into an entity.

Single record

Provide a JSON object with the field values. Template syntax is fully supported:

{
  "title": "{{ $anythink.steps.generate.data[0].title }}",
  "status": "draft",
  "author": 2
}

Bulk create

To create multiple records at once, pass an array returned by a Run Script step:

{{ $anythink.steps.parse_step.data[0].items }}

This creates one record per item in a single operation, without needing a loop.

Accessing the created record

After creating, the new record including its generated id is available to subsequent steps:

{{ $anythink.steps.create_order.data[0].id }}

Update Data

Modify an existing record.

Configuration

Example

Update the status of the record that triggered the workflow:

• Entity: orders
• Record ID: {{ $anythink.trigger.data.id }}
• Fields: {"status": "processed"}

Only the fields you specify are changed. Everything else is untouched.

Delete Data

Remove a record from an entity.

Deletions are permanent and cannot be reversed from within a workflow.

Call an API

Make an HTTP request to any external service — a webhook, a third-party API, or any other system.

Configuration

Important notes

• The URL field does not support template syntax. It must be a fixed URL. Put variable parts (IDs, slugs) in the request body instead
• Do not set Content-Type manually. It is set automatically. Including it yourself will cause the request to fail
• Headers and body fully support {{ $anythink.steps.step_name.data[0].field }} templates

Example: calling an external AI API

Headers:

{
  "x-api-key": "{{ secrets.MY_API_KEY }}",
  "anthropic-version": "2023-06-01"
}

Body:

{
  "model": "claude-haiku-4-5-20251001",
  "max_tokens": 500,
  "messages": [
    {
      "role": "user",
      "content": "Summarise: {{ $anythink.steps.get_article.data[0].content }}"
    }
  ]
}

Accessing the response

The full response body is available to downstream steps via data[0]:

{{ $anythink.steps.call_api.data[0].content[0].text }}

Run Script

Execute a JavaScript function for custom logic that cannot be expressed as a simple step.

Structure

Your script must return a value. The returned value is automatically wrapped in an array by the platform, so it is always accessed as data[0]:

return { status: "processed", count: 42 };

// Accessed downstream as:
// {{ $anythink.steps.my_script.data[0].status }}
// {{ $anythink.steps.my_script.data[0].count }}

Accessing previous step data

Previous steps data is available via $anythink:

var items = $anythink.steps.fetch_items.data;
var trigger = $anythink.trigger.data;

var results = [];
for (var i = 0; i < items.length; i++) {
  if (items[i].status === "active") {
    results.push({ id: items[i].id, name: items[i].name });
  }
}

return { items: results, count: results.length };

Important scripting notes

• Use var declarations and traditional for (var i...) loops. Modern JS features such as optional chaining and for...of may not be supported in all environments
• Never return a bare empty array. If your script might return no results, use a flag pattern so downstream steps can detect the empty case:

return {
  items: results,
  has_results: results.length > 0 ? "yes" : "no"
};

• The platform wraps your return value in an array. Even if you return { x: 1 }, it arrives downstream as [{ x: 1 }] and is accessed via data[0].x
• Strings containing newlines must be escaped before including in JSON template bodies. Build and format strings in the script, then return them ready to use

Common use cases

• Parsing and filtering items from an external API response
• Deduplicating a list of URLs or records before storing
• Building a prompt string from multiple data sources
• Computing derived values such as dates, counts, or formatted strings
• Transforming an array of items to pass to a bulk Create Data step

Condition

Branch the workflow based on the value of a field. A Condition step has two outputs: one path when conditions are met, and another when they are not.

Configuration

Condition operators

Referencing fields in conditions

For fields from a Run Script step, which always wraps output in an array:

$anythink.steps.my_script.data[0].my_field

For fields from the trigger:

$anythink.trigger.data.status

Using Condition as a gate

A common pattern is using a Condition step as a gate — only continue if a condition is true and stop the workflow otherwise. When the condition is not met, the job shows as "Condition evaluation failed" in job history. This is expected behaviour for a gate, not an error.

Dual-gate pattern

When a workflow fetches items from an external source and then filters them, you may end up with nothing at various stages. Use two Condition gates for safety:

1. gate_items — did the source return any items at all?
2. gate_results — did any items survive the filter step?

This prevents a downstream Create Data step from receiving an empty array, which can cause unexpected behaviour.