Processing data with code

Why use code to process data

Most workflows can be built visually using nodes.
However, some situations require custom logic — for example:

Calculating or transforming data dynamically
Filtering or mapping items based on conditions
Combining or formatting complex objects

In Agentic WorkFlow, you can write small pieces of Python code to handle these advanced cases inside a workflow.

The Code node

The Code node allows you to write and execute Python snippets directly inside your workflow.
It runs securely in the browser using Pyodide — a WebAssembly-based Python runtime.

You can use this node to:

Create or modify fields in your data
Validate or filter results before sending them to another node
Perform calculations or aggregations

Accessing workflow data

The Code node automatically receives all input data from the previous node(s).
You can access and manipulate it using standard Python syntax.

Example: Add a calculated field

# Access the input data
items = input_data

# Add a new field to each item
for item in items:
    value = item.get("price", 0)
    item["tax_included"] = value * 1.2  # Add 20% tax

# Return the modified data
return items

This example takes an array of objects with a price field and adds a new tax_included field to each one.

Using variables and references

You can reference values from other nodes using the standard templating syntax:

# Example using a value from another node
def compute_total():
    base = {{ $('FetchPrices').item.price }}
    return { "total": base * 1.2 }

compute_total()

When executed, this code retrieves the value from the FetchPrices node and performs a simple computation.

Best practices

When using the Code node:

Keep code small and focused on a single task
Always return data in the same structure as input (list of objects)
Use comments to explain complex logic
Test each change incrementally

If your logic becomes complex, consider breaking it into smaller workflows and use Lambda Workflows.