This page documents the original routes, which keep working. For new integrations, use the
/v1 routes in Workflows Overview. Legacy maps each original route to its /v1 replacement.Core Concepts
The Workflows API models pipelines as directed acyclic graphs (DAGs):Execution Flow
- List workflows to find the
workflow_idyou want to run. - Run the workflow with input data and receive a
job_id. - Poll live status every 10 seconds or slower until
overall_statusis no longerqueued,running, orfinalizing. - Download the result files (CSV, plus JSON or NDJSON when requested).
For workflows that start with a
webhook block, see Incoming Webhooks. For recurring runs, see Workflow Schedules.Quick start
Uploading an input file
Workflows that start with aread_csv block read their rows from a file you upload first.
multipart/form-data with a single file part. Accepts .csv, .json, .jsonl, and .ndjson files up to 45 MB. Column names cannot start with _, except _row_id.
cURL
handle_id, valid only for your organization:
handle_id as specs_override.resource_handle_id when you call /workflows/run. Runs of a read_csv workflow require it:
cURL
NDJSON input & output
Upload NDJSON or JSONL input the same way as CSV (see Uploading an input file). For NDJSON output, addresult_formats=ndjson when running. CSV is always saved; NDJSON is added alongside it. NDJSON keeps numbers, booleans, and nested objects typed, where CSV writes every value as text (lists and objects as JSON).
cURL