SillyTavern in Russia: Claude, Gemini and DeepSeek with no VPN or third-party proxy
SillyTavern (Silly Tavern, ST) is a free interface for role-play chats with characters that you run on your own computer or phone. It has no models of its own: replies come from whichever AI model you connect over an API. RuAPI is an API gateway for SillyTavern: one key opens Claude, Gemini, DeepSeek, GPT, Kimi and Qwen through the Custom (OpenAI-compatible) source. It works from Russia without a VPN, and the balance is topped up in USDT — no foreign card needed.
Updated:
from openai import OpenAI
client = OpenAI(
api_key="sk-...",
base_url="https://www.ruapi.ai/v1",
)
resp = client.chat.completions.create(
model="deepseek-flash",
messages=[{"role": "user", "content": "Hello"}],
)
print(resp.choices[0].message.content)You can test your key with this example (cURL tab) before setting up SillyTavern. In SillyTavern, the same address https://www.ruapi.ai/v1 goes into Custom Endpoint (Base URL) and the key into Custom API Key.
Why people connect SillyTavern to RuAPI
SillyTavern is only an interface; the model comes separately.
The developers state that they provide no online or hosted services: SillyTavern is installed locally and talks to external APIs or to a model on your own GPU (github.com/SillyTavern/SillyTavern, 2 Nov 2026). You cannot get your own Anthropic or OpenAI key from Russia — Russia is on neither company’s list of supported countries (anthropic.com/supported-countries, developers.openai.com/api/docs/supported-countries, 2 Nov 2026). A RuAPI account is created with email, Google or Yandex, and one key works with every model.
A “proxy” from a list is someone else’s server in the middle.
People searching for a “SillyTavern proxy” usually want an address that lets them talk to Claude or GPT. Who runs the free addresses on such lists is usually not stated, yet your whole conversation passes through their server, and access can vanish any day. SillyTavern itself warns next to the Reverse Proxy field that a proxy you do not run yourself is a privacy risk. RuAPI is an intermediary too, but with a legal entity, public prices and a privacy policy: we do not store the content of prompts or model responses, and we pass requests on to the model providers.
Long role-play chats burn through tokens.
With every new message SillyTavern sends the model the character description, the system prompt and the chat history — as much as fits into Context Size. The longer the context, the more each reply costs. Prompt caching cuts the price of the repeated part: for DeepSeek and GPT it works on its own, for Claude it takes one line in the connection settings (see prices below).
Guide: connect RuAPI to SillyTavern in 3 steps
You need SillyTavern already installed (how to set it up on Windows and Android is in the FAQ below). Menu names are as in the English interface of version 1.19.0.
Sign up for RuAPI and top up
Create an account at www.ruapi.ai with email, Google or Yandex — we do not ask for a phone number. Top-ups are in USDT, from 5 USDT (TRC20, BEP20, TON, Polygon), or in USDC; how to buy USDT with rubles via SBP is described at docs.ruapi.ai/en/buy-usdt. Foreign-issued cards work too.
A new account usually gets $0.05 to try straight away: until the first top-up claude-haiku-5-5, deepseek-flash and gpt-6-luna are available, after it every model is.
Create a separate key for SillyTavern
Console → API Keys → “Create API key”. A separate key with a spending limit suits role-play: the console log shows what each reply cost, and the limit keeps a long session from spending more than you planned.
Pick the Custom (OpenAI-compatible) source in SillyTavern
- Open API Connections (the plug icon in the top bar) and choose Chat Completion in the API list.
- Chat Completion Source → Custom (OpenAI-compatible).
- Custom Endpoint (Base URL): https://www.ruapi.ai/v1 — with /v1 and no trailing slash; SillyTavern adds /chat/completions itself. Custom API Key: your sk-… key.
- Click Connect: the Available Models list fills with RuAPI models. Pick one or type it into Enter a Model ID — for example deepseek-flash, claude-sonnet-5-5 or gemini-3.5-flash. The Test Message button sends a short check message (it also spends a few tokens).
How to test the key outside SillyTavern: docs.ruapi.ai/en/quickstart. What error codes mean: docs.ruapi.ai/en/errors
What to enter in SillyTavern
Custom (OpenAI-compatible) — for every model
Custom Endpoint (Base URL): https://www.ruapi.ai/v1
Custom API Key: sk-…
Enter a Model ID: e.g. deepseek-flash
Claude source — Claude only
Reverse Proxy → Proxy Server URL: https://www.ruapi.ai/v1
Proxy Password: sk-…
Payment
USDT / USDC from 5 USDT
foreign-issued cards
Model prices for role-play chats in SillyTavern
Models for role-play chats, from budget options to the strongest ones, in USD per 1M tokens, from the pricing page. No subscription: only the tokens of your chats are charged.
| model | input $/M | output $/M |
|---|---|---|
| deepseek-flash | $0.17 cache read $0.0120 | $1.44 |
| deepseek-v4-pro | $1.58 cache read $0.0600 | $4.75 |
| claude-haiku-5-5 | $0.12 cache read $0.0120 >100K: $0.60 | $0.60 >100K: $3.00 |
| gemini-3.5-flash | $1.80 cache read $0.18 | $10.80 |
| kimi-k3.1-pro | $2.40 cache read $0.24 | $12.00 |
| claude-sonnet-5-5 | $2.40 cache read $0.12 | $12.00 |
| claude-opus-5-5 | $4.80 cache read $0.24 | $24.00 |
input is new input tokens, “cache read” is the repeated start of a request served from the cache, output is the character’s reply. The cost of a reply grows with the context: with 30,000 tokens of context (about 1,000 of them new) and a 400-token reply it is about $0.0056 on deepseek-flash and about $0.077 on claude-sonnet-5-5 without cache, and about $0.0011 and $0.011 with cache (our estimate). For DeepSeek and GPT the cache works on its own. For Claude in the Custom source, open Additional Parameters and put the line cache_control: {type: ephemeral} into Include Body Parameters — RuAPI will place the cache mark on the last message. The line is for Claude only: remove it when you switch to a model from another vendor. A cache write costs 1.25 times the regular input price, so caching pays off while the start of the prompt stays the same; it is broken by macros such as {{random}}, lorebook entries triggered by keywords, and old messages being trimmed once the chat hits Context Size. Details: docs.ruapi.ai/en/prompt-caching
The grey “>100K” line is the vendor’s rate for long prompts: if a prompt goes over the threshold, the whole request is billed at that rate.
The RuAPI gateway or a “proxy” from a list
| RuAPI | Free proxy from a list | |
|---|---|---|
| Who is behind the service | CYBERBIT PTE. LTD. (Singapore), with published terms and privacy policy | Usually not stated |
| Your conversations | We do not store the content of prompts or model responses; requests are passed on to the model providers (docs.ruapi.ai/en/legal/privacy) | The server owner can technically read and keep everything that passes through |
| Access | Your own key, balance and request log in the console | A shared address and password that can be changed or shut down without notice |
| Models | Claude, Gemini, GPT, DeepSeek, Kimi, Qwen, GLM through official upstreams | Up to the owner; the source of the keys is usually not disclosed |
The proxy warning is in the SillyTavern 1.19.0 interface (github.com/SillyTavern/SillyTavern, checked 2 Nov 2026). On the risks of suspiciously cheap and opaque APIs: docs.ruapi.ai/en/cheap-api-risks