{
  "billing_concepts": [
    {
      "id": "bil_input_vs_output_tokens",
      "name": "Input tokens are typically billed at a different rate than output tokens",
      "category": "token_pricing",
      "description": "Most LLM API providers, including Anthropic, bill output tokens at a materially higher per-token rate than input tokens, because generating tokens is more computationally expensive than processing them as input.",
      "how_it_works": "When estimating the cost of a workflow, the split between input and output token volume matters — a request with a huge input document but a short answer costs differently than one with a short prompt and a long generated response.",
      "source_note": "General, widely-known industry pattern for token-based LLM API pricing; no specific current Anthropic rate is asserted, since this session did not perform a live pricing search.",
      "created_at": "2026-07-02 08:30:36",
      "cite_as": "https://subagentcredits.com/api/billing-concepts/bil_input_vs_output_tokens"
    },
    {
      "id": "bil_prompt_caching_cost_effect",
      "name": "Prompt caching reduces effective cost for repeated context",
      "category": "token_pricing",
      "description": "When a prompt prefix is cached (see subagentcontext.com's prompt-caching concept), reusing that cached prefix across requests is billed more cheaply than reprocessing the same tokens fresh each time.",
      "how_it_works": "This makes prompt caching a direct lever on cost for workflows that repeatedly send the same large system prompt, tool schema, or reference document alongside a smaller varying portion of the request.",
      "source_note": "Cross-references subagentcontext.com's prompt-caching concept page; general mechanism, grounded in the same prompt-caching doc mirrored in this repo (docs/docs/platform.claude.com/docs/en/build-with-claude/prompt-caching.md), no specific discount percentage asserted since not independently re-verified this session.",
      "created_at": "2026-07-02 08:30:36",
      "cite_as": "https://subagentcredits.com/api/billing-concepts/bil_prompt_caching_cost_effect"
    },
    {
      "id": "bil_token_based_pricing",
      "name": "Token-based pricing",
      "category": "token_pricing",
      "description": "The Anthropic API bills usage based on the number of tokens processed, not a flat per-request fee — both the tokens sent in a request (input) and the tokens generated in the response (output) count toward billed usage.",
      "how_it_works": "Each API call's cost is a function of input token count plus output token count, multiplied by that model's per-token rate. Different models have different rates, and larger/more capable models generally cost more per token than smaller ones.",
      "source_note": "General mechanism description; no specific dollar-per-token figures are asserted here. This session did not perform a live web search to confirm current exact pricing, so no numeric rate is stated.",
      "created_at": "2026-07-02 08:30:36",
      "cite_as": "https://subagentcredits.com/api/billing-concepts/bil_token_based_pricing"
    }
  ]
}