Frequently asked questions

codexkey is a Claude API gateway with pay-as-you-go billing that gives you access to Anthropic models. You do not need an Anthropic account: sign up, get an API key and start sending requests.

Claude models include Opus 5.5, Opus 5, Sonnet 5, Sonnet 5.5, Fable 5, Haiku 4.5 and earlier Opus and Sonnet versions. Besides Claude there are GPT-6 and GPT-5.6, Gemini, DeepSeek, Qwen, GLM, Grok, Kimi and MiniMax. Images come from GPT Image and Nano Banana, video from Veo, Kling, Seedance and others. The current list with prices is on the Models page.

No. codexkey issues the keys itself, so you need neither an Anthropic account nor a separate Anthropic key.

Retention periods and training terms are set by the API supplier and the whole chain of recipients. Checking those terms is not finished, so there is no general guarantee that data is not stored or used for training. Start with the data and recipient documents in the legal section.

Both formats work: the Anthropic Messages API and OpenAI Chat Completions. Use the one your code is already written for.

You pay per million tokens, and each model has its own rate. The full table is on the Pricing page.

No. A balance has no expiry date.

You can top up from $1. A single payment is capped at $10,000. The top-up dialog shows the exact limits for the payment method you choose.

Available methods are YooKassa (Russian bank cards, SBP), CryptoCloud (cryptocurrency), CryptoBot (Telegram) and Platega (SBP QR, cards, international cards, crypto).

Yes. You can pay in cryptocurrency through CryptoCloud and CryptoBot.

SDKs that use the OpenAI format take https://api.claudexia.tech/v1. Claude Code and the Anthropic SDK use a different address: https://api.claudexia.tech, with no /v1.

Yes, on a paid Cursor plan. Turn on the OpenAI-compatible mode, enter https://api.claudexia.tech/v1 and a Cursor model name such as cursor5. The full table of names is in the Cursor guide.

Yes. Set ANTHROPIC_BASE_URL to https://api.claudexia.tech without /v1, put the key in ANTHROPIC_AUTH_TOKEN and define the models in ANTHROPIC_DEFAULT_*. A ready-made settings.json is in the Claude Code guide.

Yes. Point base_url in the OpenAI SDK at https://api.claudexia.tech/v1 and use your codexkey API key.

codexkey keys start with sk_cdx_.

Claude Sonnet 5 is enough for everyday work: it is fast and inexpensive. When you need the best quality, take Claude Opus 5.5. In our Codex guide the default model is GPT-6 Sol.

Yes, every Claude model accepts images as input.

Extended thinking is available on Opus and Sonnet models. To turn it on, pass the thinking parameter in the request.

Most text models hold 1M tokens, including Claude Opus 5.5, Sonnet 5 and GPT-6. Claude Haiku 4.5 has a 200K context and Grok models have 256K. Exact values for each model are in the catalog.

Premium has all the models. Cheap costs less and charges one price for any token, but only some models are in it. The channel is chosen per key in the API Keys section. More on the Premium and Cheap channels page.

Yes. Images are created with POST /v1/images/generations and video with POST /v1/videos. You are charged only for a successful result. More on the Images and video page.

The provider can limit requests under account or channel rules. The dashboard key API does not set RPM, TPM or concurrency for a key, and no numeric default quotas are published. Handle HTTP 429 in your client.

Sign up on the site, then open Dashboard → API Keys to create a key.

Yes. Create as many keys as you need.

A key can have a name, a Premium or Cheap channel, a Premium fallback for Cheap and an expiry date, and its usage shows up in the statistics. RPM/TPM, concurrency, budgets, model lists, IP filters and IDE restrictions cannot be set through the dashboard key API.

Yes. Two-factor authentication through TOTP can be turned on in your account settings.

Issue a new key in the dashboard, move your applications to it, and only then delete the old one.