api
endpoint
One OpenAI-shaped endpoint
Point Claude Code, Codex, Cline, or any OpenAI client at /v1. Chat, Responses, images, audio, embeddings — translated per provider, streamed back in your format.
Cloud inference relay
One endpoint.Every model.
RelayStation routes your tools across 120+ AI providers through a single OpenAI-compatible /v1 — with fallback, OAuth refresh, and metering built in. Scoped keys and per-plan limits keep every tenant isolated.
terminal
$curl $HOST/v1/chat/completions
$-H "Authorization: Bearer rk-…"
120+providers wired
1openai endpoint
24/7managed relay
Claude✦OpenAI✦Gemini✦Codex✦Kiro✦Cursor✦Qwen✦DeepSeek✦GLM✦Kimi✦Grok✦Mistral✦Gemini CLI✦Antigravity✦OpenRouter✦Qoder✦Claude✦OpenAI✦Gemini✦Codex✦Kiro✦Cursor✦Qwen✦DeepSeek✦GLM✦Kimi✦Grok✦Mistral✦Gemini CLI✦Antigravity✦OpenRouter✦Qoder✦
Chapter 01 — Component lab
Built to route.
Everything between your prompt and the provider — handled once, in one place.
layers
fallback
Combos that never sleep
Chain models into fallback combos. When one account rate-limits or errors, the next takes over mid-flight — no retries in your code.
key
auth
OAuth vault
Claude, Codex, Gemini, Kiro, Cursor and more. Tokens refresh silently; 401s retry automatically.
bar_chart
metering
Meter everything
Per-key usage, cost estimates, and request logs out of the box.
hard_drive
isolation
Tenant-isolated by design
Scoped API keys, per-plan limits, and per-workspace usage. One tenant never sees another.
cloud_sync
regions
One endpoint, everywhere
A single managed URL for every tool and teammate — no tunnels, no port-forwarding, no VPN.
Chapter 02 — How it works
Live in three moves.
01
Connect providers
Add OAuth logins or API keys in the operator dashboard. Test each connection with one click.
02
Publish a plan
Set token, request, and RPM limits. Subscribers self-serve at /station — you never touch their machines.
03
Hand out /v1
Users mint an rk- key and call the OpenAI-compatible endpoint from any tool. You watch the meters.
Chapter 03 — Operator setup
Terminal to first token.
Two minutes from signup to first token. Create an account, mint a key, call the endpoint from anything that speaks OpenAI.
bash
$curl $HOST/v1/models -H "Authorization: Bearer rk-…"
$curl $HOST/v1/chat/completions -H "Authorization: Bearer rk-…" -d '{"model":"auto","messages":[{"role":"user","content":"hello"}]}'
Ready
Plug in. Route. Ship.
Create a station account, mint an rk- key, and call the OpenAI-compatible API.