An AI API relaybuilt for real traffic.
Extended access includes 50,000,000 free relayed tokens every UTC day at 40 RPM, backed by protected accounts, revocable keys, strict validation, and durable abuse controls.
- free tokens / UTC day
- 50.0M
- rate limit
- 40 RPM
- catalog entries
- 10
- relay status
- setup
curl https://yoinkrelay.com/v1/chat/completions \
-H "Authorization: Bearer $YOINKLAB_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"model": "llama-3.3-70b-instruct",
"messages": [
{ "role": "user", "content": "Explain SSE in one paragraph." }
]
}'capabilities
Security and function, designed together.
protocol
OpenAI-compatible requests
Use the familiar Chat Completions request shape, including multi-turn messages and server-sent event streaming.
authentication
Real accounts, no shortcuts
Argon2id password hashing, opaque server-side sessions, strict cookies, origin checks, and durable rate limits protect every account.
api keys
Keys are shown once
API keys use 256 bits of randomness. Only a server-peppered HMAC and a short display prefix are stored.
validation
Bounded and validated input
Body limits, strict parameter ranges, UTF-8 checks, and schema validation reject malformed or oversized traffic before relay work begins.
observability
Usage you can inspect
Per-account request history records model, status, latency, and token counts without storing prompt or completion content.
operations
Provider stays configurable
Connect one OpenAI-compatible upstream through server-only environment variables. Credentials never reach the browser or database.
Use it in three steps
Create an account
Register with a unique passphrase of at least twelve characters.
Create an API key
Copy it once, keep it in a secret manager, and revoke it whenever needed.
Call the endpoint
Set the relay base URL in your compatible client and send a validated request.
catalog
Configured model mappings
| model | publisher | context | status |
|---|---|---|---|
Llama 3.3 70B Instruct llama-3.3-70b-instruct | Meta | 128k | awaiting provider |
Llama 3.1 8B Instruct llama-3.1-8b-instruct | Meta | 128k | awaiting provider |
Qwen2.5 72B Instruct qwen-2.5-72b-instruct | Alibaba | 32k | awaiting provider |
Qwen2.5 Coder 32B qwen-2.5-coder-32b | Alibaba | 32k | awaiting provider |
Mistral Small 3.1 mistral-small-3.1 | Mistral AI | 94k | awaiting provider |
faq
Straight answers
Is the relay connected to an AI provider?+
The status shown on this page is authoritative. Until the owner adds an upstream provider key, chat requests return a clear 503 instead of a simulated answer.
Are raw passwords or API keys stored?+
No. Passwords are hashed with Argon2id. API keys are generated from 256 random bits, shown once, and stored only as a keyed HMAC.
What does request logging contain?+
The application records operational metadata such as model, status, latency, and token counts. It does not persist prompt or completion text.
Can the upstream provider still charge for usage?+
Yes. This deployment does not charge a relay software fee, but whichever AI provider the owner connects may bill under its own terms.
Does this guarantee zero vulnerabilities?+
No responsible operator can promise that. The service uses layered controls, pinned dependencies, automated updates, least privilege, and repeatable security checks to reduce risk.
Create a protected account and issue your first key.
The key is displayed once and can be revoked instantly from the dashboard.
Create account