generated: '2026-08-14'
method: searched
source: >-
  https://docs.parallel.ai/findall-api/features/findall-preview +
  https://parallel.ai/pricing + https://docs.parallel.ai/integrations/mcp/search-mcp
notes: >-
  Parallel publishes NO separate sandbox environment. There is no test host, no
  test-mode key prefix, no test/live key split, no fixtures and no time
  simulation — every call goes to production api.parallel.ai against the live
  web, which is inherent to a web-research API. What Parallel does publish
  instead is a set of low-cost and no-cost ways to try a call before committing:
  hosted playgrounds, a FindAll `preview` generator, an anonymous free MCP
  endpoint, and monthly free credits. Those are captured below verbatim. Nothing
  here is invented; absent fields are recorded as absent.
separate_sandbox_environment: false
test_mode: false
key_prefixes:
  test: null
  live: null
  note: Single key type, sent as x-api-key. No test/live distinction is documented.
test_clocks: false
fixtures: false
test_data: null
trial_surfaces:
- kind: playground
  name: Task playground
  url: https://platform.parallel.ai/play/task
  http_status: 200
  auth: platform account
- kind: playground
  name: Search playground
  url: https://platform.parallel.ai/play/search
  http_status: 200
  auth: platform account
- kind: playground
  name: Extract playground
  url: https://platform.parallel.ai/play/extract
  auth: platform account
- kind: playground
  name: Monitor playground
  url: https://platform.parallel.ai/play/monitor
  auth: platform account
- kind: playground
  name: Find All playground
  url: https://platform.parallel.ai/play/find-all
  auth: platform account
- kind: anonymous-endpoint
  name: Search MCP (free tier)
  url: https://search.parallel.ai/mcp
  auth: none
  note: >-
    Callable with no account and no API key — the closest thing Parallel has to a
    zero-friction try-before-you-buy surface. Verified live 2026-08-14.
  docs: https://docs.parallel.ai/integrations/mcp/search-mcp
cheap_preview_modes:
- product: FindAll
  parameter: 'generator: "preview"'
  match_limit: 5-10
  cost: $0.10 fixed, $0.00 per match
  purpose: >-
    Validate query structure, match conditions and enrichment schema on a small
    candidate sample before committing to a full run. Same endpoint as a real run.
  docs: https://docs.parallel.ai/findall-api/features/findall-preview
free_allowance:
  signup_credit: Up to $80 at signup
  monthly_credit: $5 in free credits per month
  free_requests: Up to 5,000 requests per month
  source: https://parallel.ai/pricing
billing_safety:
  failed_runs_billed: false
  note: Task API charges only successfully completed runs; Responses API charges only successful responses.