OpenAI-compatible inference

One endpoint for every model.

Point Claude Code, Codex, Cline or any OpenAI client at a single /v1 URL. Start free, and pay per token on private capacity when you need more.

No credit card. Out of tokens? Watch an ad for more.

request

$curl $HOST/v1/chat/completions \

$ -H "Authorization: Bearer sk-…" \

$ -d '{"model":"r/claude-sonnet-5-5",

$ "messages":[{"role":"user","content":"hello"}]}'

response200 · streaming

{ "model": "r/claude-sonnet-5-5",
  "choices": [{ "message": {
    "role": "assistant",
    "content": "Hello! How can I help?" } }] }

Model families

Why RelayStation

Built to stay out of your way.

A relay should be invisible. It routes, meters and recovers, so you can focus on what you ship.

api
Drop-in OpenAI compatibility
Chat, Responses, images, audio and embeddings behind one /v1. Change the base URL and key, keep your code.
layers
Automatic fallback
When a model is rate-limited or errors, the next one takes over mid-flight. No retry logic in your app.
bolt
Never blocked on free
Out of free tokens? Watch an ad for another million, up to ten times a day. Nobody gets a hard stop.
lock
Private paid capacity
Paid plans run on a dedicated pool, kept separate from free traffic, billed per million tokens.
bar_chart
Metered per key
Usage, spend and request logs for every API key, visible in your account.

How it works

01
Create an account
Sign in with GitHub. You start on the free plan instantly, with no card.
02
Mint an API key
Generate an sk- key in your station. Revoke or reveal it any time.
03
Call /v1
Use the key from any OpenAI-compatible tool. Usage shows up in your account right away.

Pricing

Free

No card needed

Pay per token

Per M tokens

Private, dedicated capacity