Get your first response in 3 minutes
One key, one endpoint, every model. The API is OpenAI-compatible, so your existing SDK works. You pay per token from a prepaid balance funded with USDT on TON.
Add funds
Open the dashboard, tap Add funds and send USDT on the TON network. The minimum is $2. Your balance appears after the transaction is confirmed on-chain, usually under a minute.
Create a key
Go to API in the dashboard and tap Create key. The key starts withtg_live_ and is shown once. Store it somewhere safe and never share it. You can revoke it at any time and create another.
Send a request
Replace tg_live_YOUR_KEY with your key and run this.
curl https://toolgram.app/api/v1/chat/completions \
-H "Authorization: Bearer tg_live_YOUR_KEY" \
-H "Content-Type: application/json" \
-d '{
"model": "openai-gpt-oss-20b",
"messages": [{"role": "user", "content": "Hello!"}],
"max_tokens": 100
}'What you just called
One base URL serves every model: https://toolgram.app/api/v1. Paths and request bodies follow the OpenAI format, so anything that speaks OpenAI can be pointed here.
| Method | Path | What it does |
|---|---|---|
| POST | /chat/completions | Chat and tool calling. Streaming with stream: true |
| POST | /completions | Classic text completion |
| POST | /embeddings | Vectors for search and clustering |
| GET | /models | Every model id available to your key |
Keep your existing SDK
Change two lines in any OpenAI client: the base URL and the key. Model names are passed through, so use the ids from the table below or from Models and prices.
Good models to start with
Prices are per 1 million tokens. Cheapest input first, live from the catalog.
Next
Ask in the group with your key prefix, the model id, the error and the time. We answer there. Never send the full key.