Skip to content
QUICKFLO DOCS

Run an Action for Each CSV Row

The most common CSV workflow: take a file of records and do one thing per row. Parse the CSV into rows with data.csv-to-items, loop them with core.for-each, and put the per-row action inside the loop. The example logs each row so it runs the moment you paste it — swap the log for the real action and the shape doesn’t change, whether you’re calling an API, writing to a data store, or sending an email.

Workflow JSON

Section titled “Workflow JSON”

The recipe takes the CSV inline so it runs as-is. To feed it from an uploaded file instead, Setup shows how to read the CSV from a form file upload.

csv-for-each-row.json
{
  "name": "Run an Action for Each CSV Row",
  "initial": {
    "csv": "email,firstName,plan\njane@acme.com,Jane,pro\nbob@acme.com,Bob,starter\ncarol@acme.com,Carol,pro"
  },
  "steps": [
    {
      "stepId": "parse-csv",
      "stepType": "data.csv-to-items",
      "input": {
        "source": "content",
        "csv": "{{ initial.csv }}",
        "skipHeader": true
      }
    },
    {
      "stepId": "process-rows",
      "stepType": "core.for-each",
      "input": {
        "items": "{{ parse-csv.items }}",
        "concurrency": 8,
        "continueOnError": true,
        "steps": [
          {
            "stepId": "process-row",
            "stepType": "core.log",
            "input": {
              "message": "Processing {{ $item.email }} ({{ $item.firstName }}) on the {{ $item.plan }} plan. Replace this step with the real action: an HTTP call, a data store write, an email."
            }
          }
        ]
      }
    },
    {
      "stepId": "respond",
      "stepType": "core.return",
      "input": {
        "formSubmission": {
          "type": "success",
          "message": "Processed {{ process-rows.count }} rows",
          "data": {
            "Total": "{{ process-rows.count }}",
            "Succeeded": "{{ process-rows.succeeded }}",
            "Failed": "{{ process-rows.failed }}"
          }
        }
      }
    }
  ]
}

Setup

Section titled “Setup”

Trigger: add a Form trigger with a file upload field so a partner can drop a CSV and the workflow runs on submit. The uploaded file arrives at {{ initial.<field>.url }} — point the parse step at it by switching source to file and setting fileUrl:

{
  "stepId": "parse-csv",
  "stepType": "data.csv-to-items",
  "input": {
    "source": "file",
    "fileUrl": "{{ initial.leads.url }}",
    "skipHeader": true
  }
}

Reading from a file URL streams the rows instead of holding them in memory, so one run can iterate a file far larger than it could load at once. For files over ~30 MB, raise the field’s Max file size in the form builder.

How it works

Section titled “How it works”
CSV (inline, or a form file upload)
  → parse into rows
  → for each row, concurrency 8:
      → run the action (the example logs the row)
  → return how many succeeded and failed

concurrency controls how many rows run at once — 8 here. Raise it to move faster, lower it to stay under an API’s rate limit. continueOnError keeps the batch going when a single row fails; the failure is counted, not fatal. core.for-each reports count, succeeded, and failed, which the Return step hands back as the submission summary.

What to customize first

Section titled “What to customize first”
  • Swap the action. Replace the process-row log step with whatever you need per row — a data store write, an email, an LLM call, or a sub-workflow. Everything inside steps runs once per row with {{ $item }} bound to that row.
  • Tune concurrency to the downstream limit. If the API you’re calling allows 10 requests/second, keep concurrency low and add a Rate Limit check inside the loop rather than cranking it up.
  • Collect the failures. Each row’s result is in {{ process-rows.items }}. Add a step after the loop that filters to the failed ones and writes them to a data store or emails them back, so a partner knows which rows didn’t go through.
  • Validate before you act. Add a core.if step at the top of the loop that skips rows missing a required field, so a malformed line doesn’t burn an API call.
Section titled “Related recipes”