One API key for Claude, GPT and Gemini. One dashboard that controls every app using it — models, prompts, budgets — with zero redeploys.
Sign in, create a key, paste it into your bot, site or automation. Every major model answers behind it.
claude, gpt, gemini — pick per request, switch with a word. A new model drops? It's already behind your key.
No Anthropic or OpenAI account yet? Create a key and make your first call right away — plug in your own provider keys later, when you want direct billing.
Telegram & Discord bots, websites, Zapier / Make / n8n, scripts — anything that takes an OpenAI-style key takes tinyai.
Set a monthly cap — $5 means $5, then requests stop. Every call listed in plain dollars.
You already use AI APIs. The pain isn't calling them — it's maintaining five projects while models, prompts and prices keep changing. Move the AI layer here and stop redeploying.
Your apps point here once. Everything you'd normally hardcode — model, prompt, limits — becomes a dashboard setting.
Your apps point at tinyai once. After that, changing the model is a dashboard click — every project follows instantly. No edits, no releases, no "which repo was that in?"
Edit the system prompt, reshape requests, transform data in and out — centrally, per project — instead of hunting through source code you shipped months ago.
Plug in your Anthropic / OpenAI / Google keys: every request is billed by the provider to you, at their prices — you don't pay us. Keys stored encrypted. no markup
A key per project: see exactly who spends what, cap each one, revoke in a click. Your whole AI operation on one screen.
Mint a key for a teammate, an assistant, a student — capped at a few dollars a month, cheap models only if you like, revocable anytime. They build; you keep control of the spend.
tinyai itself costs nothing right now. You only ever pay the AI providers — and only for what you actually use.
Any key, off, one click — without touching your provider accounts or breaking anything else.
tinyai speaks OpenAI's /v1/chat/completions — requests,
responses and streaming. If your code speaks OpenAI, it already speaks tinyai:
the only change is the base URL.
# The only change to your code: the base URL. from openai import OpenAI client = OpenAI( base_url="https://ai.tinyappmart.com/v1", api_key="sk-...", # one tinyai key ) client.chat.completions.create( model="claude-opus-4-8", # or haiku, sonnet, gpt, gemini... messages=[{"role": "user", "content": "Hello!"}], )
Define a custom endpoint at /v1/x/<name> — declarative
field mapping or sandboxed JS. Your format in, your format out.
Non-OpenAI-compatible API? Write a small outbound adapter and tinyai translates on the way out.
Server-sent events exactly like OpenAI's — token by token, same frames,
same [DONE].
The always-current model list lives in your dashboard.
tinyai is free while in preview. The models themselves are billed by the providers at their own prices — through your keys, or through a capped key someone shared with you.
Yes — we have to, to sign your requests. They're kept encrypted on our server, decrypted only in memory at the moment a request is sent, and never shown in any UI or returned by any API — the dashboard displays only the last 4 characters.
Nothing. We route requests; providers bill you directly. No markup, no per-token fee.
tinyai speaks the standard OpenAI API. Point your base URL back at any provider and everything keeps working. Zero lock-in — which is exactly why it's safe to try.