What Is api.foo? One API Key for Claude, GPT, Gemini, Grok, DeepSeek and More
On this page
In one sentence: api.foo is an LLM API gateway. Create one API key in the console and call Claude, GPT, Gemini, Grok, DeepSeek, Kimi, GLM, MiniMax and more through a single base URL, https://api.foo, with pay-as-you-go billing in Chinese yuan.
About the name: api.foo is not a placeholder
Programmers often use "foo" as a stand-in name in sample code, so api.foo can look like a fake address. It is a real, operating website. .foo is a regular top-level domain run by Google, just like .dev or .app.
- Website and console: https://api.foo
- API base URL:
https://api.foo(https://api.foo/v1also works for OpenAI-style clients) - Docs: https://api.foo/docs/en/
What it does
- One key, many model families: Claude, GPT (including Codex and GPT Image), Gemini (including Nano Banana), Grok, DeepSeek, Kimi, GLM, MiniMax and more. The model square lists over 70 models, each with its own description and pricing page.
- Three API formats: OpenAI (
/v1/chat/completions,/v1/responses,/v1/images/generations), Anthropic (/v1/messages) and Gemini (/v1beta). Point your existing client at the new base URL and key. - Works with common tools: Claude Code, Codex, Gemini CLI, OpenCode, Cursor, Cline, Cherry Studio, CC Switch and TRAE. See the docs for step-by-step setup.
- Every charge is itemized: the console's usage log shows the model, input / output / cache tokens and cost of each request.
Billing and payment
| Item | Details |
|---|---|
| Currency | Chinese yuan. A ¥10 top-up gives a ¥10 balance; no dollar conversion |
| Charging | Per request by actual usage (image models per image); no monthly fee |
| Minimum top-up | ¥5; ¥40 when paying with USDT |
| Payment methods | Alipay, WeChat Pay, bank card, USDT (BNB Smart Chain / BEP20 only) |
| Prices | See the model square; the live page is authoritative |
Keys are tied to a "group": you pick one group when creating a key, and the key can call the models in that group. One account can have several keys sharing one balance.
How it relates to the official APIs
api.foo is an independently operated third-party service. It has no affiliation or partnership with Anthropic, OpenAI, Google, xAI or other model makers; model names and trademarks belong to their owners.
Your client sends the request to api.foo, api.foo forwards it to an upstream model service, and the result comes back unchanged. That makes it a fit for:
- developers who want to pay with Alipay or WeChat Pay and settle in yuan;
- people who want one key for several model families instead of separate accounts and cards with each vendor;
- users of Claude Code, Codex, Cursor and similar tools who need a working custom endpoint.
Upstream rate limits or overloaded models can still cause short-lived errors, so set up retries, budget caps and a fallback for production workloads. For what to check before choosing any gateway, see OpenAI API alternatives.
How data is handled
Per the privacy policy:
- Conversation content is not stored: prompts, context, attachments and model output only pass through the servers while being forwarded and are not written to the database or logs.
- Only billing metadata is recorded: time, model name, the name of the key used, input and output token counts, cost, latency and source IP.
- Exception: queued asynchronous jobs such as video or music generation keep the prompt and the result link so you can check progress and fetch the output later.
Whatever service you use, keep passwords, keys, ID numbers, card numbers and trade secrets out of your prompts.
Refunds
Per the terms of service:
- Reviewed and fixed: duplicate charges, a successful payment that did not reach your balance, and billing errors confirmed by the platform. These are refunded to the original payment method or credited back.
- No unconditional refunds: normal usage caused by misconfiguration, client mistakes, leaked keys or going over budget.
- How to ask: email [email protected] with your account, order number, request time and screenshots.
Get started in three steps
- Sign up at api.foo, open Console → API Keys and create a key.
- Top up in Console → Wallet, from ¥5.
- Put the base URL
https://api.fooand your key into your tool. For long outputs and large requests, use the direct linehttps://dc.api.foo(API only, no web pages).
Contact
- Email: [email protected] (best for orders, charges and refunds, so there is a written record)
- Telegram group: t.me/api_foo
- QQ group: 639791134; WeChat and QQ support QR codes are on the about page
Include your sign-up email, order number or request time, the exact error and screenshots. Never send your full API key; the first few characters are enough to identify it.
Related
FAQ
Is api.foo a real website?
Yes. api.foo is an operating LLM API service, not a placeholder domain from code samples (programmers often use "foo" as a stand-in name). The website and the API base URL are both https://api.foo.
Is api.foo official or affiliated with OpenAI, Anthropic or Google?
No. api.foo is an independently operated third-party service with no affiliation or partnership with Anthropic, OpenAI, Google, xAI or other model makers. It forwards your requests to upstream model services and returns the results unchanged.
Which models can I use through api.foo?
Claude, GPT (including Codex and image models), Gemini (including Nano Banana image models), Grok, DeepSeek, Kimi, GLM, MiniMax and more, over 70 models in total. The model square at api.foo/pricing shows the current list and prices.
How does api.foo charge?
Prepaid balance in Chinese yuan, charged per request by actual token usage (image models per image), with no monthly fee. Top-ups start at ¥5 (¥40 for USDT) and can be paid with Alipay, WeChat Pay, bank card or USDT on BNB Smart Chain.
Does api.foo store my prompts and responses?
According to its privacy policy, no. Prompts, context, attachments and model output only pass through its servers while being forwarded and are not written to its database or logs. Only billing metadata such as time, model, token counts and cost is recorded. Queued asynchronous jobs such as video or music generation are the exception.
Which tools work with api.foo?
Claude Code, Codex, Gemini CLI, OpenCode, Cursor, Cline, Cherry Studio, CC Switch, TRAE and any program that speaks the OpenAI, Anthropic (Claude) or Gemini API format. The docs have a setup guide for each tool.
Found a mistake? Email [email protected].