CSV Tools

Parse, transform, and generate CSV data.

Overview

The CSV Tools connector is a built-in utility for working with CSV data inside a workflow. It parses incoming files into structured rows, transforms them (filter, sort, dedupe, add columns, merge), and serializes the result back to CSV for downstream steps like FTP uploads, email attachments, or imports into another system.

Use it as the interchange format when one side of a pipeline only speaks CSV - FTP drops, finance system exports, legacy back-office uploads - or as a staging shape between an API source and a relational destination. CSV Tools is the heart of most ETL workflows on the platform.

What You Can Do

The CSV Tools connector exposes these tools:

parse - Parse a CSV string or file into an array of row objects.
from-json - Convert an array of JSON objects into CSV.
to-json - Convert CSV into a JSON array (alias for parse when you want JSON out).
filter - Keep rows matching a condition.
sort - Sort rows by one or more columns.
dedupe - Remove duplicate rows (by all columns or a subset).
merge - Combine multiple CSV files or strings into one.
add-column - Add a new column with a constant or computed value.
transform - Apply a row-level transform to reshape columns.
slice - Take a row range from a larger CSV.
info - Return metadata: row count, column names, file size, encoding.

Authentication and Setup

No connection or authentication is required. These tools are built into the platform and available in every workflow by default - just drop a Connector node onto the canvas and pick the tool you need.

Using in a Workflow

Add a Connector node, select CSV Tools, and pick a mode:

Direct Mode - Best for ETL pipelines where the steps are fixed. Chain parse -> filter -> transform -> from-json in a deterministic sequence.
Agent Mode - Useful when the AI needs to inspect a CSV (info, parse) before deciding how to clean it.

For very large CSVs, use slice with a Loop node to process in chunks - the connector handles streaming well, but downstream AI or HTTP steps may have payload limits.

Tips

Run info first. Get the column names and row count before parsing - lets the workflow branch on shape.
Dedupe before write. Run dedupe on import workflows to avoid creating duplicate records downstream.
Validate then filter. Pair CSV Tools with Validation Tools - filter out rows that fail validation before pushing to the destination.
Use add-column for sync markers. Stamp imported_at or source_workflow_id onto each row so downstream debugging is easier.
Prefer from-json over hand-building CSV strings. Quote escaping is a constant source of bugs - let the tool handle it.

Common Pitfalls

Delimiter assumptions. European CSVs often use ; as the delimiter and , as the decimal separator. Set the delimiter explicitly on parse.
Encoding. Files from Windows or older finance systems may be Latin-1 or UTF-16. If you see mojibake, override the encoding on parse.
Headers vs no headers. A CSV with no header row will parse to numerically-keyed columns. Always confirm the source format.
Numeric coercion. Columns like ZIP codes and product SKUs with leading zeros lose them when treated as numbers. Force string types.
Memory on huge files. A 500MB CSV won't fit in a single workflow step's memory. Slice it.

Common Use Cases

For technical API details and field specifications, see the CSV Tools documentation.