The unified global LLM API gateway
Compatible with OpenAI and Anthropic protocols. Instant top-up, no VPN, no subscription. Change one base_url — zero business code changes.
$ curl https://www.ruapi.ai/v1/chat/completions \
-H "Authorization: Bearer sk-..." \
-H "Content-Type: application/json" \
-d '{
"model": "claude-sonnet-5",
"messages": [{"role": "user", "content": "Hello"}]
}'Only BASE_URL and KEY — zero business code changes.
RuAPI in numbers
Production-grade infrastructure
Three reasons teams across Russia and the CIS run RuAPI in production.
Direct access from Russia
Multi-region Cloudflare edge with a Yandex Cloud egress route. Requests go through without a VPN and are not blocked by ISPs.
Original models, no downgrade
Official upstreams only, no silent swaps to cheaper models. Genuine models — what you see is what you get, transparent pricing.
Transparent USDT / USDC payments
USDT / USDC in 5 minutes on TRC20, BEP20 or Polygon, gas under $1, fully on-chain. Balance never expires; unused funds are refundable within 7 days.
Connect in a minute
Enter the RuAPI URL and key in the tool you already use — no code changes. Each tool has a step-by-step guide.
Claude Code
Anthropic's official command-line coding assistant. No Anthropic subscription or account needed.
export ANTHROPIC_BASE_URL="https://www.ruapi.ai" export ANTHROPIC_AUTH_TOKEN="sk-..." claudeStep-by-step guide →
Cursor
AI code editor. In Cursor Settings → Models, fill in two fields:
OpenAI API Key: sk-... Override OpenAI Base URL: https://www.ruapi.ai/v1Step-by-step guide →
Codex CLI
OpenAI's official command-line coding assistant; works with GPT models. Add to ~/.codex/config.toml:
model_provider = "ruapi" model = "gpt-5.4" [model_providers.ruapi] name = "RuAPI" base_url = "https://www.ruapi.ai/v1" env_key = "RUAPI_API_KEY" wire_api = "responses"Step-by-step guide →
Cline
An AI coding agent for VS Code. Set API Provider to OpenAI Compatible, then fill in:
Base URL: https://www.ruapi.ai/v1 API Key: sk-... Model ID: claude-sonnet-5Step-by-step guide →
n8n
Workflow automation. Create an OpenAI credential (Credentials → New → OpenAI) and fill in:
API Key: sk-... Base URL: https://www.ruapi.ai/v1Step-by-step guide →
OpenAI / Anthropic SDK
Your own code: change only base_url; the rest stays the same.
OpenAI SDK: base_url = "https://www.ruapi.ai/v1" Anthropic SDK: base_url = "https://www.ruapi.ai"Step-by-step guide →
One KEY, every leading model
OpenAI and Anthropic protocol compatible. 50+ models supported.
# Generated automatically from the pricing table, updated October 2, 2026 · full list →
5 minutes from $0 → sk-…
No card verification, no waiting list, no KYC.
Create account
Email OAuth · grab your sk-* key in 30 seconds
Top up
Auto-credited · no credit card, no SIM verification
Set BASE_URL
One env variable, every OpenAI / Anthropic SDK works · zero changes in your business code
Original models, no swapping. Transparent pricing.
Per-token billing. No monthly fee, no minimum spend, no dynamic pricing.· live
| model | input $/M | output $/M |
|---|---|---|
| claude-opus-5-5 | $4.8 est. effective w/ cache ≈ $1.152 | $24 |
| claude-sonnet-5 | $2.4 est. effective w/ cache ≈ $0.672 | $12 |
| gpt-5.5 | $6 est. effective w/ cache ≈ $1.68 | $36 |
| gpt-5.4-mini | $0.9 est. effective w/ cache ≈ $0.252 | $5.4 |
| gemini-3.1-pro-preview | $2.4 est. effective w/ cache ≈ $0.672 | $14.4 |
| deepseek-v4.1-flash | $0.168 est. effective w/ cache ≈ $0.043 | $1.44 |
| deepseek-v4-pro | $1.584 est. effective w/ cache ≈ $0.365 | $4.752 |
| glm-5.3 | $1.44 est. effective w/ cache ≈ $0.576 | $5.016 |
Prices are per 1M tokens; the input column shows the regular (uncached) input price. The “est. effective w/ cache” line is an estimate at an 80% input-cache hit rate — it is not a discount, just a calculation for agent workloads. The price includes: payment in USDT/USDC without a foreign card, the native Anthropic API, one key for all models, support in Russian, and a refund of any unused balance within 7 days.
Why it costs less in practice
Cached input is billed at a discount
For coding assistants and agents, repeated context (system prompt, open files, chat history) is often 80–97% of the input. Whatever hits the cache is billed at a discount: for Claude Sonnet 5, a cache read costs a tenth of the regular input price. On the pricing page, models that support caching show a small-print "est. effective w/ cache" figure — an estimate, not a discount.
A real bill: the same day, about 24× cheaper
A real one-day Claude Code bill: 405M tokens, $167.24 on Claude Sonnet 4.6. The same tokens cost $6.99 on DeepSeek V4.1 Flash and $37.10 on DeepSeek V4 Pro.
No hidden fees
You pay only for what you use: no monthly fee, no minimum spend, no long-context surcharge and no dynamic pricing; the balance never expires. Each request's model, input and output tokens, and charge are shown in the usage log.
RuAPI vs. the alternatives
The common ways to use foreign LLM APIs from Russia.
| RuAPI | Direct (OpenAI / Anthropic) | OpenRouter | ProxyAPI | |
|---|---|---|---|---|
| Payment | USDT / USDC or Visa / Mastercard, from $5 | Needs a foreign card | Card, USDC (USDT only via workarounds) | Russian cards, SBP, rubles |
| Works from Russia without a VPN | Yes | Usually needs a VPN (geo-restrictions) | No, returns 403 (checked 30 Sep 2026) | Yes |
| Models on one key | Claude, GPT, Gemini, DeepSeek, Qwen, GLM and more — 50+ | Only that provider's models | Many providers | Many providers |
| Native Anthropic API (Claude Code) | Yes, works with DeepSeek too | Anthropic: yes, Claude only | Yes, but officially Anthropic models only | Yes, works with DeepSeek too |
| Model precision | Original models, no quantization | Original models | Some providers serve FP8 / INT4 | — |
| Support | In Russian, via Telegram | In English | In English | In Russian |
Information about other services is taken from their websites (checked September 2026) and may change. Detailed comparisons: RuAPI vs OpenRouter · RuAPI vs ProxyAPI