Changelog
Recent updates to apiToken.sale — new Claude and GPT models, guides, localization, and platform improvements.
2026-09Daily and monthly spend caps, model allowlist, customer Activity API
Each customer key can now take an optional UTC-day period spend cap, UTC-month period spend cap, and a model allowlist from the published catalog, on top of the lifetime spending limit and expiration date. Authenticated customers can query GET /activity for request-level tokens, charged nanoUSD, public errors, and agent session grouping.
2026-09Public error envelopes match each official API
Every supported lane now returns the official error envelope of that protocol: Anthropic JSON (HTTP 402 is billing_error), OpenAI {error:{message,type,param,code}} (prepaid money stays HTTP 402 insufficient_quota, never 429), and Google {error:{code,message,status,details}} (invalid keys remain 400 API_KEY_INVALID; money is 402 FAILED_PRECONDITION). The KIMI lane on kimi/* uses that same Anthropic envelope; the extra 400s are documented_limitation for mcp_servers and provider-executed tools, plus unsupported_parameter for count_tokens. A present official field this endpoint cannot honour is fail-closed 400 documented_limitation or unsupported_parameter with a named field and a workaround — never a silent 200. Exact strings: /docs/errors.
2026-08One unified router endpoint for every provider
https://router.apitoken.sale now serves all three providers on a single endpoint: the native Anthropic lane (POST /v1/messages), the OpenAI lanes (POST /v1/responses and the OpenAI-compatible universal lane POST /v1/chat/completions for any catalog model) and the native Gemini lane (/v1beta/models/{model}:generateContent) — all with the same sk-pool key. A unified namespaced catalog (anthropic/*, openai/*, google/*) is available at GET /v1/models. The legacy per-provider hosts (api.apitoken.sale, openai.api.apitoken.sale/v1, gemini.api.apitoken.sale) remain fully supported for existing integrations.
2026-07One flat 50% discount — tiers retired
B2C pricing is now a flat 50% off official provider spend on every request, for every account and any top-up amount. The Starter/Builder/Pro/Studio/Scale ladder, cumulative top-up thresholds and 30-day tier retention are gone — there is nothing to unlock and nothing to keep.
2026-07GPT-5.6 on the same key — apiToken.sale is now multi-provider
The OpenAI-compatible API at openai.api.apitoken.sale/v1 serves the GPT-5.6 line (sol, terra, luna), gpt-5.5 and gpt-5.4 through Responses and Chat Completions with SSE streaming. One sk-pool key and one prepaid balance cover both Claude and GPT.
2026-07Tier ladder rebalanced
B2C tier discounts now progress from 60% to 70% off official spend (Starter 60%, Builder 62.5%, Pro 65%, Studio 67.5%, Scale 70%). Thresholds and 30-day holds are unchanged; Starter stays free at 60%.
2026-07Guides in Russian and Chinese
The full Claude API guide library is now available in English, Russian and Simplified Chinese with per-language URLs.
2026-07Guide library expanded
Added new guides covering gateways, rate limits, streaming, prompt caching, key security, and building AI agents on Claude.
2026-06Claude Opus 4.8 available
Opus 4.8 is available on the same API key and prepaid balance as Opus 4.7, Sonnet 5, Sonnet 4.6 and Haiku 4.5.
2026-07Per-key lifetime guardrails
Set an optional lifetime spending limit and expiration date when you create a key, then update or remove either guardrail from the dashboard.
2026-05Progressive discount tiers
B2C accounts start 60% below official spend and progress up to 70% off as cumulative top-ups grow.