Pay as you go
No packages or prepaid plans. Each model has a price per 1M tokens.
For people who build
Your editor, agent or product.
Claude, GPT and Gemini.
One connection between them.
Your side
The connection
1API
One keyShared balance
OpenAI- and Anthropic-compatible interfaces.
Named keys, a shared balance and usage statistics.
07Questions
An API gateway to Claude, GPT and models from other developers. The same models, the familiar Anthropic and OpenAI formats, pay as you go with no monthly billing.
It will if your tool lets you set a custom OpenAI- or Anthropic-compatible API address and your own key. Cursor, Claude Code, VS Code extensions, OpenCode and any SDK with a base URL setting work this way. Check the model and features you need with a test request.
Both: the Anthropic Messages API and the OpenAI Chat Completions API.
Opus flagships, balanced Sonnet models and GPT models, all behind stable IDs. The lineup follows the provider, so check the catalog or the /v1/models response for the current list and prices.
In the catalog, for the selected model and channel. Input and output tokens are priced differently. The calculator on this page estimates spend from the live rates.
No. You pay as you go, per 1M tokens for each model, with no monthly billing.
By card, SBP (Russia's Faster Payments System), MTS wallet, YooKassa, Platega or crypto (USDT, BTC, ETH via CryptoCloud and CryptoBot). A foreign card is not required.
No. Requests go through the codexkey gateway at our API address, so no VPN is needed. You never hold an Anthropic account, so there is nothing to ban: you work with your own codexkey key and balance.
No. The account has one shared balance. Named keys show how usage is split across your tasks.
No. Your balance has no expiry date.
Unused paid advances can be refunded under the refund rules in the legal section. Incorrect charges are corrected. Promotional credit is tracked separately and cannot be exchanged for cash. Contact support or the legal email in the operator details.
Join our Telegram channel and get $1 on your balance. That is enough to run real requests, measure latency and confirm your tooling works. If you then decide to top up, the minimum is also $1.
Requests are sent to an external API provider. Its retention and processing terms are its own, and there is no promise that data is not stored or used for training. Read the legal section before sending data. Authenticator-based 2FA is available for sign-in.
One address accepts Anthropic Messages and OpenAI Chat Completions. A key has a name, a Premium or Cheap channel, optional Premium fallback for Cheap, an expiry date and usage statistics. The dashboard key API does not set per-key RPM/TPM, concurrency, budgets, model allowlists, IP filters or IDE restrictions.
There is nothing to ban: you hold no Anthropic account. Requests go through the codexkey gateway, and you work with your own codexkey key and balance.
An aggregator such as codexkey puts Claude, GPT, Gemini, DeepSeek and other models behind one API key, one balance and one request format, so you do not need a separate account with every developer.
Sign up for codexkey, top up your balance, and create a key in the dashboard. The key works with the Anthropic Messages format and with the OpenAI-compatible API, and you pay per token.
Create a key in the codexkey dashboard and top up the shared balance. GPT models in the catalog use the same key, including through the OpenAI-compatible format.
Gemini is part of the codexkey catalog and uses the same key, balance and payment methods as the other models. The models page lists the current versions and the price per token.
Top up the codexkey balance by SBP, a Russian card or crypto, then pick a DeepSeek model in the catalog. Usage is charged per token from the shared balance, with no subscription.
05Features
No subscriptions, no rate-limit games. Tokens in, tokens out.
No packages or prepaid plans. Each model has a price per 1M tokens.
Works with SDKs, editors, agents and automation that speak the Anthropic or OpenAI format.
Join our Telegram channel and get a dollar on your balance. Enough to check latency and compatibility before you pay anything.
Requests pass through our API supplier. Where and how data is processed is described in the legal section. Do not send personal or confidential data before you have reviewed those terms.
Legal section01Models
One API key and one catalog: Claude, GPT, Gemini, DeepSeek and more. Check what each model can do and what a token costs before your first request.
| Family | Models in the catalog | Input, per 1M tokens |
|---|---|---|
| Claude12 models | Claude Opus 5.5, Claude Opus 5, Claude Fable 5.1 and 9 more | from $0.08 |
| GPT10 models | GPT-6 Astra, GPT-6.1 Sol, GPT-6 Sol and 7 more | from $0.04 |
| Gemini8 models | Gemini 3.8 Flash, Gemini 3.7 Flash, Gemini 3.1 Pro and 5 more | from $0.2 |
| Qwen5 models | Qwen 3.8 Max, Qwen 3.8 Flash, Qwen 3.8 Omni Flash and 2 more | from $0.1 |
| DeepSeek5 models | DeepSeek V4.1 Flash, DeepSeek V4 Pro, DeepSeek V4 Flash and 2 more | from $0 |
| GLM3 models | GLM 5.3, GLM 5.3 Flash, GLM 5.2 | from $0.06 |
| MiniMax3 models | MiniMax M3, Hailuo 02, MiniMax M3.1 Flash | from $0.06 |
| Kimi3 models | Kimi K2.7 Code, Kimi K3, Kimi K3.1 Pro | from $0.42 |
| Hunyuan1 model | Hunyuan 4 Preview | from $0.12 |
| Grok4 models | Grok 4.7, Grok 4.6, Grok 4.5 and 1 more | from $0.28 |
| Seed3 models | Seedance 2.5, Seedance 2.0, Seedance 1 | |
| Kling2 models | Kling 3.0, Kling 2.6 | |
| PixVerse1 model | PixVerse V6 | |
| Runway2 models | Runway Gen-4.5, Runway Gen-4 |
02Pricing
You pay for tokens, not for a subscription: every model has its own price per 1M. Below are the live rates for both channels, Premium and Cheap, and a spend calculator.
Pick a model for the job: the table shows what each model is good at, what it can do and what it costs.
Direct access to the models, every feature, very cheap caching. Usually the cheapest for IDE work.
Scroll the table sideways to see every column
Estimate your spend before you top up
200K in, 30K out, on Claude Fable 5
Presets are typical volumes, not a promise. Put your own numbers in: rates come from the live price list.
03Usage
Create named keys for teams, projects and integrations, and use per-key statistics for usage accounting.
Every key draws from the shared balance. There is no separate cap per team or key, and keys do not get independent quotas.
Premium or Cheap channel, optional Premium fallback for Cheap, an expiry date and usage per key.
04Connection
ANTHROPIC_BASE_URL=https://api.claudexia.tech
ANTHROPIC_AUTH_TOKEN=sk_cdx_YOUR_KEY
OPENAI_BASE_URL=https://api.claudexia.tech/v1
OPENAI_API_KEY=sk_cdx_YOUR_KEYYou get the key in the dashboard. Model IDs are in the catalog and in the /v1/models response.
06Blog
Plus any tool that lets you set a custom OpenAI- or Anthropic-compatible API address.