Pay per token
Input, output, cache reads and cache writes are billed separately, each at its own price. No subscription and no monthly minimum. A request that is refused or fails before generation starts costs nothing.
One key to Claude, GPT and Gemini for Claude Code and Codex
AI Creator API sells developers tokens for Claude, GPT, Gemini and other models. One key works in Claude Code, Codex, Cline, OpenCode and Aider, and you pay only for tokens.
Claude Opus 5.5: $0.248 per 1M input tokens and $1.24 per 1M output. The official Anthropic price is $4 and $20. Other models
Three steps
to the first request- 01
Create a key
Sign in to the dashboard and create a key on the Keys page. It is shown once, so save it right away.
- 02
Top up
On the Billing page: by card in rubles or by crypto, an amount from $1 to $1000.
- 03
Paste the config
Change the base URL and the key in your tool. The ready configs are below.
Set the variables in your terminal and run claude.
unset ANTHROPIC_API_KEY
export ANTHROPIC_BASE_URL=https://api.aicreator.app
export ANTHROPIC_AUTH_TOKEN=sk-aic-…
export ANTHROPIC_DEFAULT_OPUS_MODEL=claude-opus-5-5
export ANTHROPIC_DEFAULT_SONNET_MODEL=claude-sonnet-5-5
export ANTHROPIC_DEFAULT_HAIKU_MODEL=claude-haiku-5-5
claudeAdd the provider to config.toml and put the key in AIC_API_KEY.
model = "gpt-6-astra"
model_provider = "aic"
[model_providers.aic]
name = "AI Creator"
base_url = "https://api.aicreator.app/v1"
env_key = "AIC_API_KEY"
wire_api = "responses"
stream_idle_timeout_ms = 300000Any OpenAI Chat Completions client needs three fields: the base URL, the key and the model’s full name. This is how they look in Cline.
Base URL: https://api.aicreator.app/v1
API Key: sk-aic-…
Model: gpt-6-astraPrices per 1M tokens
in / outA few models from the price list. Our price comes from our price list, the official one from the vendor’s pricing page.
| Model | With us | Official | Saving |
|---|---|---|---|
| Claude Opus 5.5Anthropic | $0.248 / $1.24 | $4 / $20 | −93% |
| Claude Sonnet 5.5Anthropic | $0.147 / $0.733 | $2 / $10 | −92% |
| Claude Fable 5.1Anthropic | $0.411 / $2.055 | $10 / $50 | −95% |
| GPT-6 AstraOpenAI | $0.198 / $0.99 | $10 / $50 | −98% |
| GPT-6.1 SolOpenAI | $0.094 / $0.469 | $2 / $10 | −95% |
| Gemini 4 Argon*Google | $0.643 / $2.546 | $2 / $10 | −67% |
- Our prices are as of ; the official ones are from the pricing pages of Anthropic, OpenAI, Google, checked on November 9, 2026.
- The saving is the smallest of three — on input, on output and on cache reads — rounded down. Not every model on the list is cheaper than the official price; the full table shows which.
- * From December 1, 2026 Gemini 4 Argon officially costs $4 / $20.
Official Anthropic APIClaude Opus 5.5
- Input · 200K
- $0.80
- From cache · 800K
- $0.16
- Output · 100K
- $2.00
Total$2.96
AI Creator APIClaude Opus 5.5
- Input · 200K
- $0.050
- From cache · 800K
- $0.010
- Output · 100K
- $0.124
Total$0.184
16× cheaper
How billing works
A balance in dollars
You top up in advance. Before a request we reserve an estimated amount, after the answer we charge by the actual tokens, and the rest of the reserve returns to the balance at once.
Top up from $1
By card in rubles or by crypto, an amount from $1 to $1000. With a card the ruble amount is computed at the exchange rate at the time of payment, and the balance gets exactly the dollars you asked for. With crypto the payment service may refuse an amount below its own minimum.
What the dashboard gives you
Usage
- Spend, requests, tokens and errors by day and by model.
- A request log: model, key, protocol, input, output and cache tokens, price, latency, HTTP status and error code.
- Filters by key and model, export to CSV or JSON.
Keys and limits
- Up to 20 active keys per account. A separate key per tool or project shows who spends what.
- A key can carry its own limits: requests and tokens per minute, a monthly spend cap in dollars and a list of allowed models.
- Limits are set per key. A key can be revoked at any time.
What to know up front
- Who runs it
- AI Creator API is a product of AI Creator, the image, video and music generation service.
- Support
- Support chat on Telegram
- Data
- We don’t store the text of prompts and answers, only request metadata: token counts, cost, status and latency.
- Limits and errors
- Service limits, key limits and error codes are in the docs, with what to do about each error.
Questions
Which models are available?
32 models from OpenAI, Anthropic, Google, Moonshot, DeepSeek, Zhipu and xAI. The list with prices is on the pricing page; the models your key can use come from GET /v1/models.
Which protocols does the API speak?
Anthropic Messages (/v1/messages) for Claude Code, OpenAI Responses (/v1/responses) for Codex CLI and OpenAI Chat Completions (/v1/chat/completions) for Cline, OpenCode, Aider and the OpenAI SDK. Not every model is on every protocol: see the Protocols column of the pricing page.
How do I set up Claude Code?
Set ANTHROPIC_BASE_URL to https://api.aicreator.app, without /v1, and ANTHROPIC_AUTH_TOKEN to your key. Drop ANTHROPIC_API_KEY from the environment and pin the models behind the names Opus, Sonnet and Haiku; the ready block is above on this page. Details are in the Claude Code guide.
How do I pay from Russia?
By card in rubles or by crypto, an amount from $1 to $1000. With a card the ruble amount is computed at the exchange rate at the time of payment, and the balance gets exactly the dollars you asked for. With crypto the payment service may refuse an amount below its own minimum.
The API is unreachable from Russia. What do I do?
If https://api.aicreator.app is unreachable or answers get cut off, use the same API at https://api-ru.aicreator.app: the key, the paths and the balance are the same.
What are the limits?
Per key, at most 120 requests and 2,000,000 input tokens per minute (tokens are estimated from the request body size) and 4 requests at a time. From one IP address, at most 60 requests per minute, 300 per five minutes and 1,200 per hour to each API endpoint; for one machine this limit is stricter than the key’s. A request body of up to 32 MB, and one request lasts 30 minutes at most. A key can carry its own limits below these. The full list is in Errors and limits.
What does a failed request cost?
A request that is refused or fails before generation starts costs nothing: the whole reserve returns to the balance. If generation breaks off midway, you pay for the input and for what the model managed to generate; if the token counts had not arrived yet, input and output are estimated. More in Billing and usage.
Do you store my prompts?
We don’t store the text of prompts and answers, only request metadata: token counts, cost, status and latency.
