generated: '2026-09-03'
method: searched
source: https://cracked.ai/docs/how-it-works (Playground section) + https://cracked.ai/llms.txt
model: no-key playground plus trial credit (no separate sandbox environment; test-vs-live is a single live environment with free tiers)
playground:
  endpoint: POST /v1/try {"provider","endpoint","input"}
  allowance: $0.05 of tool credit per address per day, at most 20 tries
  eligibility: synchronous tools from live providers estimated at $0.02 or less per call; file, messaging and async tools excluded
  allowance_check: GET /v1/try
  cli: npx cracked-ai try "weather in austin"
  ui: every tool page has a Run button that works signed out
free_endpoints_no_key:
  - POST /v1/discover
  - POST /v1/inspect
  - GET /v1/capabilities
  - GET /v1/x402
  - GET /v1/bounties/program
trial_credit:
  human_signup: $2 after email confirmation
  agent_self_registration: $1 via POST /v1/agents/register (5 per address per day); claim by a human adds $1
key_prefixes:
  live: ck_live_
notes: no test-mode key prefix or simulated data is published; the playground bills a shared platform workspace