BetterToken
Works in Russia without a VPN · Pay in RUB

One reliable API for all your AI toolsLeading AI models—up to 32% less than official prices

One API for OpenAI, Anthropic, and Gemini models—with automatic retries and transparent request costs.

Works with
Claude CodeCodexOpenCoden8nCursorDifyOpenClawClineRoo CodeContinueAider
and more
Models & pricing
Claude
GPT
GeminiComing soon
DeepSeekComing soon
GLM
QwenComing soon
Kimi
MiniMaxComing soon
Connect in 3 steps

Connect your AI tool and send your first request

Everything stays on one screen: sign up, create a key, and copy the ready-to-use configuration.

Selected

Claude Code

CompatibleGuided setupAnthropic

Sign in or create an account

Sign in or register and get $3 for your first test request.

Create an API key

Open the form, choose a name and group, and your key will be ready to connect.

Configure the tool and test a request

Restart Claude Code, run `claude` in your project folder, and send a short test task.

Commandcurl -fsSL https://bettertoken.ai/install-claudecode-provider.sh | bash
Claude Code setup guide
Do not see your tool? Any client that accepts a custom API key and compatible API URL can connect.Open documentation
Models & pricing

Compare pricing and availability across 4 providers

Anthropic

Claude

7-day availability from the BetterToken catalog
claude-sonnet-5
Input₽106.76/M
Output₽533.8/M
7-day availability
97.0%
claude-sonnet-4-6
Input₽160.14/M
Output₽800.7/M
7-day availability
97.7%
claude-opus-5
Input₽266.9/M
Output₽1334.5/M
7-day availability
97.5%
claude-opus-4-8
Input₽266.9/M
Output₽1334.5/M
7-day availability
97.9%
claude-opus-4-7
Input₽266.9/M
Output₽1334.5/M
7-day availability
100.0%
claude-opus-4-6
Input₽266.9/M
Output₽1334.5/M
7-day availability
100.0%
claude-haiku-4-5-20251001
Input₽53.38/M
Output₽266.9/M
7-day availability
98.5%
claude-fable-5
Input₽533.8/M
Output₽2669/M
7-day availability
97.6%
from 68% of official pricingOpen the full pricing catalog
Transparent billing

Transparent costs for every API request

Billing example

Pay only for what you use

Your final charge is calculated from the listed model price and actual input and output tokens.

Model price
Listed
Input tokens
Itemized
Output tokens
Itemized
Final chargeActual usage
No hidden multipliers.
Billing example

Every charge is traceable

Model, time, token counts, cache usage, final cost, and status stay together in one request record.

Model & time
Logged
Tokens & cache
Itemized
Cost & status
Recorded
Request logFull breakdown
All billing details stay with the request.
Billing example

No charge for technical failures

If a technical failure produces no billable output, the pre-charge is returned automatically.

Pre-charge
Recorded
Billable output
None
Returned amount
Recorded
Final chargeNo charge
Valid output and user cancellation follow the refund rules.
Valid output and user cancellation follow the refund rules.Refund rules
Frequently asked questions

Everything you need before your first request

Clear answers about compatibility, errors, keys, payments, and refunds.

Yes, if it accepts a custom API key and an API URL compatible with OpenAI, Anthropic, or Gemini. The homepage lists common examples. Use the corresponding protocol guide for other tools.

BetterToken first attempts to recover the request automatically. If the error continues, check the selected model status. For a 401, verify the API key and Base URL first. For a 429, 500/503, or timeout, retry or temporarily use a recommended model.

One account can use every published model. The API key and Base URL still depend on the protocol: Anthropic-compatible and OpenAI-compatible tools may require different keys. We show the correct setup after you choose a tool.

BetterToken is available in Russia without a VPN and supports RUB payments. The top-up page always shows the currently available methods and minimum amount. If a payment is marked successful but the balance does not update, do not pay again—send support the time, amount, and record status.

Open Workspace and copy the key again. If you think it may have leaked, disable or delete it immediately and create a new one. Never send a complete API key to support or a public chat.

For a technical failure that produces no billable result, the pre-charge is returned under the refund rules. A request that returned useful output, was canceled by the user, or violated service rules may be handled differently. The final result is visible in request logs.

Start with the guide for your tool and check the model status. If the issue remains, contact support with the time, tool, model, error code, and payment record status. Do not send a full API key, prompt, or model response.

Start with one successful request

Connect BetterToken to your AI tool

Get $3 in test credit, choose your tool, and copy a ready-to-use setup.

No subscription · Pay in RUB · Costs stay visible