generated: '2026-08-14'
method: searched
source: >-
  https://docs.refuel.ai/quickstart, https://docs.refuel.ai/catalog/introduction,
  https://docs.refuel.ai/guides/applications/using_applications
description: >-
  Refuel's test surface is an in-product PLAYGROUND, not a sandbox. There is no test mode,
  no test/live key separation, no fixture or trigger tooling, and no magic test values.
  Every call an agent or developer makes runs against production data with a production API
  key. What Refuel does provide is a browser playground per deployed application and per
  catalog transformation, from which you can copy working code snippets.
test_live_separation:
  supported: false
  key_prefixes: none
  note: >-
    One API key per team, issued in the app. No `test_`/`live_` prefix, no mode header, no
    separate sandbox host. cloud-api.refuel.ai is the only base URL Refuel publishes.
playground:
- name: Application Playground
  type: hosted-console
  url: https://app.refuel.ai
  auth: account login required
  docs: https://docs.refuel.ai/quickstart
  description: >-
    Every deployed application has a Playground tab where you submit sample rows and see the
    transformed output. Refuel's quickstart calls it "a great way to test out your
    application, and retrieve the relevant code snippets to integrate it programmatically
    into your stack."
- name: Catalog transformation playground
  type: hosted-console
  url: https://app.refuel.ai/catalog
  auth: account login required
  docs: https://docs.refuel.ai/catalog/introduction
  description: >-
    "Each transformation published in the catalog has a dedicated playground, where you can
    test the transformation with your own data." You can also call the same transformation
    over REST — the docs publish
    `POST https://cloud-api.refuel.ai/applications/resume-parsing/label`.
sample_data:
- name: Yelp reviews sample dataset
  type: sample-dataset
  url: https://gist.github.com/nihit/74abe1d926acba2d45c68631f216b096
  format: csv
  description: >-
    The CSV of Yelp business reviews Refuel's quickstart tells you to upload as your first
    dataset. Published by a Refuel co-founder as a public gist. This is the only concrete
    test fixture Refuel publishes.
test_values: []
test_clocks:
  supported: false
fixtures_or_triggers:
  supported: false
gaps:
- No test mode and no sandbox environment — all testing consumes production quota.
- No published test credentials, no anonymous demo endpoint.
- >-
  The playground requires an account, and account signup is a sales-gated form at
  https://www.refuel.ai/get-started, so the console is not reachable for evaluation without
  contacting Refuel.