WEL quickstart ​

WEL (Workato Expression Language) is Workato's transformation language. It runs alongside the rest of your recipe, on Workato's data plane, with nothing to install and nothing to connect to.

This page uses one example to showcase three methods of working with WEL:

• In a recipe: Get started quickly by creating recipe datapills with the Workato Expression Language connector.
• With an AI client: Write WEL with an external AI client.
• With the binary: Test changes to an expression quickly and without running a recipe.

Example ​

A webhook delivers an order. The destination system expects a flat record with a normalized name and email, a calendar date, and an exact total.

Input

{
  "order_id": "so-1001",
  "placed_at": "2026-03-15T09:30:00Z",
  "customer": {
    "first_name": "kalani",
    "last_name": "park",
    "email": " [email protected] "
  },
  "line_items": [
    {"sku": "WID-1", "qty": 3, "unit_price": "12.50"},
    {"sku": "GAD-7", "qty": 1, "unit_price": "99.00"}
  ]
}

Formula

{
  order_id: upper(_.order_id),
  customer_name: f"{capitalize(_.customer.first_name)} {capitalize(_.customer.last_name)}",
  email: lower(trim(_.customer.email)),
  order_date: PlainDate(DateTime(_.placed_at)),
  item_count: count(_.line_items),
  total: _.line_items >> map_by(li ~> Decimal(li.unit_price) * li.qty) >> sum
}

This example uses the following components:

• _ is the input variable, which contains the preceding payload. _.order_id reads the order_id field from it. Refer to Input variables for details.
• >> pipes a value into a function, so the total reads left to right. The pipeline takes the line items, turns each into a price, and adds them up.
• Decimal(li.unit_price) keeps the money exact. The total is 136.50, not 136.5. The scale survives because Decimal is an exact type rather than a display format.
• PlainDate(...) produces a calendar date with no time zone, so it can't shift a day on its way to the destination.

Output

{
  "order_id": "SO-1001",
  "customer_name": "Kalani Park",
  "email": "[email protected]",
  "order_date": "2026-03-15",
  "item_count": 2,
  "total": 136.50
}

Start in a recipe ​

Recipes are the most common way to work with WEL. This connector doesn't require connection setup. Create a recipe or open an existing one, then complete the following steps to run the preceding example:

{
  "order_id": "so-1001",
  "placed_at": "2026-03-15T09:30:00Z",
  "customer": {
    "first_name": "kalani",
    "last_name": "park",
    "email": " [email protected] "
  },
  "line_items": [
    {"sku": "WID-1", "qty": 3, "unit_price": "12.50"},
    {"sku": "GAD-7", "qty": 1, "unit_price": "99.00"}
  ]
}

{
  "order_id": "SO-1001",
  "customer_name": "Kalani Park",
  "email": "[email protected]",
  "order_date": "2026-03-15",
  "item_count": 2,
  "total": 136.50
}

{
  order_id: upper(_.order_id),
  customer_name: f"{capitalize(_.customer.first_name)} {capitalize(_.customer.last_name)}",
  email: lower(trim(_.customer.email)),
  order_date: PlainDate(DateTime(_.placed_at)),
  item_count: count(_.line_items),
  total: _.line_items >> map_by(li ~> Decimal(li.unit_price) * li.qty) >> sum
}

Start with an external AI client ​

An external AI client that hasn't seen WEL will write something that looks plausible and doesn't run, usually by borrowing lambda syntax from another language. WEL uses ~> for a lambda, and -> or => are parse errors.

The WEL kit fixes that by providing the AI client a language reference and platform binary. This allows the AI to check generated code against the real engine and create an expression that reliably works.

An AI client with the WEL kit can create a transformation from your plain language description. Include the sample payload and what the destination needs. Ask it to verify the expression against the engine and return the sample output and formula that you can paste directly into the action.

Start with the binary ​

A recipe run takes minutes to test an expression. The wel binary runs the engine locally instead, and answers in under a second. This is a better workflow for quickly iterating while you're still writing the expression.

Complete the following steps to work with the wel binary for the provided example:

{
  order_id: upper(_.order_id),
  customer_name: f"{capitalize(_.customer.first_name)} {capitalize(_.customer.last_name)}",
  email: lower(trim(_.customer.email)),
  order_date: PlainDate(DateTime(_.placed_at)),
  item_count: count(_.line_items),
  total: _.line_items >> map_by(li ~> Decimal(li.unit_price) * li.qty) >> sum
}

{
  "order_id": "so-1001",
  "placed_at": "2026-03-15T09:30:00Z",
  "customer": {
    "first_name": "kalani",
    "last_name": "park",
    "email": " [email protected] "
  },
  "line_items": [
    {"sku": "WID-1", "qty": 3, "unit_price": "12.50"},
    {"sku": "GAD-7", "qty": 1, "unit_price": "99.00"}
  ]
}

wel eval -f order.wel -I order.json

{order_id: "SO-1001", customer_name: "Kalani Park", email: "[email protected]", order_date: 2026-03-15, item_count: 2, total: 136.50}

Useful flags ​

The wel binary accepts additional flags for testing and validation:

The wel binary has other flags and subcommands beyond these two, but they target developers building the host application that embeds the WEL engine, not recipe builders testing an expression.

Next steps ​

• Transform data: The WEL connector integration and what schema inference doesn't catch.
• Standard library: Information about every WEL function.
• Data types: Why Decimal and PlainDate were the right choices in the preceding example.
• Cookbook: Examples of common integration problems.
• Error codes: Troubleshoot job failures.