RelayAI
RelayAI
RelayAI reads your message, sends it to whichever model handles it best, and streams the answer back. When a model runs out of free calls for the day, it writes a briefing for the next one — so the conversation keeps going instead of starting over.
Free to use. You bring your own API keys — RelayAI never charges you and never sees your bill.
Five steps, all of them invisible.
Free tiers run out. Conversations shouldn't.
“Widget is allergic to chicken. The script is feed_log.py. They want answers
under 40 words. Next step: name the timestamp column.”
Reads the briefing, keeps your tone and your constraints, and answers the question you actually asked — without you repeating a word of it.
and it names who should take over nextHand off again and the third model gets every briefing, not just the newest — because summarising a summary loses detail.
A live count of what each key has left today, and how long until it resets — in the provider's own timezone, not yours.
Every account brings its own keys. Nobody else can spend your quota, and RelayAI has no bill to pass on to you.
Add more than one key per provider. When one is spent or rate-limited it is benched and the next takes over.
Attach a picture and the router automatically sends it to a model that can actually see.
Auto-routing is the default, not a cage. Pin any chat to a specific model from a dropdown.
A web app that works on your phone. Add it to your home screen and it behaves like an app.
RelayAI speaks to these today. The free ones cost nothing to try.
No — and nothing can. Those logins are locked to their own domains. RelayAI uses API keys, which are a separate product from the free web chats: a free ChatGPT account does not include API access. Google, Groq, Cerebras, Mistral and OpenRouter all hand out free API keys; OpenAI and Anthropic are pay-as-you-go only.
RelayAI itself is free and there is nothing to pay it. You bring your own keys, so whatever your providers charge — usually nothing, on their free tiers — is between you and them.
Your chats and keys are stored on the server this runs on so they are there when you come back. Keys are shown to you masked after you save them, and are never written into a file the server would execute. Your messages are sent to whichever provider answers them, under that provider's own terms — so don't paste anything into RelayAI you wouldn't paste into their web app.
It tells you plainly, and the Quota page shows exactly what is spent and when each provider rolls over. Adding a second provider is usually a two-minute fix — that is the whole reason RelayAI speaks to so many.
Every answer is labelled with the model that wrote it and why it was chosen. If you disagree, pin the chat to whichever model you prefer.
Make an account, paste in one free API key, and start talking.
Create a free account