{
  "billing_concepts": [
    {
      "id": "bil_credit_consumption",
      "name": "How prepaid credits are consumed",
      "category": "prepaid_credits",
      "description": "Credits are consumed continuously as API calls are made — each call's token-based cost (input + output tokens at that model's rate) is deducted from the account's remaining credit balance in near-real-time.",
      "how_it_works": "Because consumption is metered per-call rather than in large discrete steps, an account's remaining balance reflects actual usage closely, which is why usage dashboards typically show near-real-time (not batch-delayed) balance figures.",
      "source_note": "General mechanism description, consistent with a standard prepaid-credit billing model; no specific numeric example given.",
      "created_at": "2026-07-02 08:30:40",
      "cite_as": "https://subagentcredits.com/api/billing-concepts/bil_credit_consumption"
    },
    {
      "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_monitoring_usage",
      "name": "Monitoring usage and spend",
      "category": "usage_tiers",
      "description": "The Anthropic Console provides usage and cost visibility so an organization can track token consumption, spend against its prepaid balance, and rate-limit headroom over time.",
      "how_it_works": "Regularly reviewing usage dashboards helps catch unexpectedly high-cost workflows (e.g. a loop that re-sends a large document as fresh input on every call instead of using prompt caching) before they exhaust a credit balance or approach a rate limit.",
      "source_note": "General practice description; no specific dashboard feature names or numeric figures asserted, since this session did not independently verify current Console UI details via a live fetch.",
      "created_at": "2026-07-02 08:30:46",
      "cite_as": "https://subagentcredits.com/api/billing-concepts/bil_monitoring_usage"
    },
    {
      "id": "bil_prepaid_credits",
      "name": "Prepaid credits on the Anthropic Console",
      "category": "prepaid_credits",
      "description": "API usage on the Anthropic Console is typically funded by prepaid credits — an organization adds credits to its account balance, and API usage is deducted from that balance as requests are made.",
      "how_it_works": "When the credit balance is exhausted, further API requests are blocked until more credits are added (or, if configured, an auto-reload mechanism tops up the balance automatically once it falls below a threshold).",
      "source_note": "General billing mechanism description for a prepaid-credit model; no specific credit-package dollar amounts are stated, since this session did not perform a live pricing search to confirm current figures.",
      "created_at": "2026-07-02 08:30:40",
      "cite_as": "https://subagentcredits.com/api/billing-concepts/bil_prepaid_credits"
    },
    {
      "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_rate_limits_scope",
      "name": "Rate limits apply per organization and per model",
      "category": "usage_tiers",
      "description": "Rate limits are generally enforced per organization (not per individual API key) and can vary by model — a higher-capability model may have lower throughput limits than a smaller, faster model even within the same usage tier.",
      "how_it_works": "This means a workload spread across multiple API keys within the same organization still shares one pool of rate limit headroom, and switching to a different model within a request pattern can change which limit actually governs throughput.",
      "source_note": "General mechanism description consistent with typical LLM API provider rate-limiting design; no specific numeric limits are asserted.",
      "created_at": "2026-07-02 08:30:46",
      "cite_as": "https://subagentcredits.com/api/billing-concepts/bil_rate_limits_scope"
    },
    {
      "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"
    },
    {
      "id": "bil_usage_tiers",
      "name": "Usage tiers scale rate limits with usage history",
      "category": "usage_tiers",
      "description": "API accounts are commonly placed into usage tiers that determine their rate limits (requests per minute, tokens per minute, etc.) — higher tiers generally unlock higher throughput.",
      "how_it_works": "An account's tier is often based on its spend history and how long it has been actively using the API, rather than being fixed permanently at signup — accounts can move to a higher tier over time as usage and payment history accumulate.",
      "source_note": "General mechanism description; no specific tier thresholds, exact rate-limit numbers, or dollar-spend cutoffs are stated, since this session did not perform a live web search to confirm current exact figures for Anthropic's usage tiers.",
      "created_at": "2026-07-02 08:30:46",
      "cite_as": "https://subagentcredits.com/api/billing-concepts/bil_usage_tiers"
    }
  ]
}