Claude + GPT · Available in China

Claude at 30% of official. GPT as low as 1/30.

SACTL is a multi-model API gateway built for the China market. Claude (Opus 4.7 / Sonnet 4.6 / Haiku 4.5) at × 0.30 official; GPT (codex / gpt-5.5) as low as 1/30 of official. Point base_url at https://api.sactl.ai and you're done — Claude Code and Codex CLI both work. WeChat / Alipay / USDT / wire transfer. Telegram support. Compatible with Claude & OpenAI APIs.

Claude save 70% · GPT save ~97%
China-direct · no VPN required
Drop-in for Claude Code · Codex CLI · Cursor · SDK
curl — /v1/chat/completions
# Flip base_url — Claude via /v1/messages, GPT/codex via /v1/responses
curl -X POST https://api.sactl.ai/v1/chat/completions \
  -H "Authorization: Bearer sk-xa-..." \
  -H "Content-Type: application/json" \
  -d '{"model":"claude-opus-4-7","messages":[...]}'

# Response headers report exact cost and remaining budget
HTTP/2 200
x-sactl-usage-prompt-tokens: 248
x-sactl-usage-completion-tokens: 612
x-sactl-usage-cost-usd: 0.0143   # official $0.0478, Claude 3折 (×0.30)
x-sactl-budget-remaining-usd: 452.71

# GPT / codex CLI: base_url=…/v1, wire_api=responses, as low as 1/30
What we solve

The China-Claude friction, all of it, handled.

No overseas credit card. No proxy gymnastics. No mystery API outages. No paying full official price. Pick the answer to all of these and you've found us.

Claude 30%, GPT 1/30

Claude is × 0.30 official: Opus output $75/MTok → we charge $22.50/MTok; a $1500 monthly bill becomes $450 — $1050 saved. GPT (codex / gpt-5.5) is as low as 1/30 of official. Official up, we follow up; official down, we follow down. The ratio doesn't move.

China-direct, no VPN

No overseas VPS, no proxy stack, no ladder. Singapore + Hong Kong dual-route, ~80ms average latency from major Chinese cities. Streaming is stable; Claude Code holds long sessions without dropping.

Zero client code changes

Claude Code, Cursor, Cline, Anthropic SDK go to /v1/messages; Codex CLI and OpenAI SDK go to /v1/responses. Change one base_url and ship — code stays put.

Multi-account pool, auto-failover

An upstream account hits a risk control, a rate limit, or a 5xx? Backend swaps to the next account inside 30 seconds — your client still sees a 200. Maintaining accounts is our problem, not yours.

Pay easy, get a real invoice

WeChat, Alipay, USDT, wire transfer — all four. Top-ups land instantly, billing is per-token. Wire-transfer is supported above $1000/mo, with Chinese 普票 or 专票 issued for company expense reimbursement.

Telegram support (CN/EN)

No English ticket queue. @ us in the Telegram group; ~5-minute response during working hours. Account setup, rate limits, billing questions — answered in real time. No "consultation fee".

Why us

Calling official direct vs. through SACTL

Same Claude / GPT underneath. The difference is how the request gets there, how you pay, and who picks up when something breaks.

Dimension
Direct to Claude
Through SACTL
Claude price (Opus 4.7, $/MTok in/out)
$15 / $75 full
$4.50 / $22.50 (-70%)
GPT price (codex / gpt-5.5)
Official full
As low as 1/30
Access from China
VPN / overseas server required
Direct, ~80ms
Payment methods
Overseas card / Stripe
WeChat / Alipay / USDT / wire
Invoicing
No Chinese fapiao
普票 / 专票 on demand
Client compatibility
Each vendor's own SDK
Claude Code / Codex CLI / Cursor / SDK
Support
English ticket, queue
Telegram CN/EN · ~5 min
Account risk-control
Blocked = business halt
Multi-account auto-failover
Minimum top-up
$5 + overseas card barrier
Any amount
How it works

Three steps. First request in 5 minutes.

01

Top up via Telegram

Message our Telegram support; pay via WeChat, Alipay, USDT, or wire transfer. No minimum — $10 works. The amount lands as real balance, billed per-token, no plan lock-in.

02

Get an sk-xa-* Virtual Key

Once funded, support hands you the key directly, or you generate one in the dashboard. Set a monthly budget cap per key — leaks stop themselves instead of draining your balance.

03

Change one base_url

Point your Claude Code / Codex CLI / SDK base_url at https://api.sactl.ai, swap the API key to sk-xa-*. Code stays the same. From the next request onward you're paying 30% for Claude, as little as 1/30 for GPT.

Real numbers

Not a marketing deck. The monthly bill.

The four numbers below are what the service actually does. We won't quote "99.99%"-style theatre.

× 0.30

Discount across the Claude lineup — Opus / Sonnet / Haiku, same ratio

~80 ms

Average latency from major Chinese cities (SG/HK dual-route)

30 min

From top-up to first successful request

1/30

GPT (codex / gpt-5.5) — as low as this fraction of official

Pricing

Claude 30%, GPT 1/30, billed per-token

No monthly fee, no minimum top-up. Top up what you want, use what you need. When official prices change, ours track them — the ratio stays. See full pricing →

WeChat / Alipay

¥ scan to pay

Solo developers, side projects, small tools — fastest path to start.

  • Telegram payment slip; scan = instant credit
  • $10 minimum, no floor
  • Refundable balance
Top up via Telegram
Recommended

USDT

USDT on-chain

Technical teams, overseas projects, FX-sensitive customers.

  • 1:1 USD settlement, no FX spread
  • TRC20 / ERC20 supported
  • $500+ top-ups get priority support
  • No Chinese bank account required
Get USDT address

Wire transfer

¥ $1000+/mo

Companies needing fapiao, monthly settlement, dedicated rep.

  • Company-to-company wire, T+1 settlement
  • 普票 / 专票 issued on demand
  • Monthly settlement, annual frame contract negotiable
  • Dedicated account manager + WeChat group
Discuss wire
Boundaries, stated

Who SACTL fits, and when not to use it

No marketing-speak. This section tells you the cases where SACTL is the wrong answer — so you don't pay first and regret later.

✓ Fits

  • Developers, technical teams, and overseas-project teams calling Claude API from inside China
  • Claude Code / Codex CLI / Cursor / Cline / OpenAI SDK / LangChain users
  • Customers paying via WeChat, Alipay, USDT, or wire transfer
  • Chinese companies that need 普票 (general) or 专票 (special VAT) invoices
  • Price-sensitive medium-tail usage in the $50–$50,000/mo range
  • Anyone who'd rather not maintain Claude accounts and chase risk-control bans themselves

✗ Doesn't fit

  • Finance / healthcare / government customers with strict data-residency requirements (use self-hosted or sign with Anthropic directly)
  • Large enterprises requiring binding SLA contracts (we currently operate via Telegram + monthly settlement)
  • Users who specifically need Anthropic-account features (Workbench, Claude.ai shared quota, etc.)
  • Heavy offline-eval / huge-batch users who'd benefit from Anthropic's official 50% Batch discount (their Batch tier may be cheaper — happy to advise)

Up and running in minutes. Claude at 30%, GPT at 1/30.

Message us on Telegram with "trial". You'll get the top-up link, your API key, and the Claude Code / Codex CLI integration docs. Post in the group when something's off — ~5 minute response during working hours.