FAQ

Integration

Do I need to change code to use the platform?

No rewrite. The platform is OpenAI-compatible — the official openai SDK only needs a base_url change; Claude / Gemini use their official SDKs pointed at the platform. No code at all? Use Client Integration.

Does the base URL need /v1?

• Calling the API directly or via the OpenAI SDK: yes, include /v1 (e.g. https://gateway.mindproxy.ai/v1).
• Some desktop clients append /v1 automatically, so you only enter the domain there.
• A 404 almost always means /v1 is one too many or too few — adjust it.

Which models are supported?

OpenAI, Claude, Gemini and other mainstream models. The actual list depends on your key — see GET /v1/models or the console "Model Prices".

Can I use Claude / Gemini native formats?

Yes. Claude via /v1/messages, Gemini via generateContent. You can also call every model through the OpenAI-compatible format.

Billing

How does pricing work?

Prepaid wallet: recharge USDT, then pay by real usage. Text by token, images by image, tasks by settlement. See Recharge & Billing.

How long does a recharge take to credit?

Automatically credited after on-chain confirmation. Check order status under console "Recharge Records".

Will a dropped streaming request overcharge me?

No deliberate overcharge. The platform settles on the real usage produced as best it can; when real usage isn't available it falls back to the estimate, and the excess hold is released by a background safety net.

Are failed tasks charged?

No. Failed / cancelled / expired tasks release the hold and aren't charged.

Are cache hits cheaper?

Yes. On a prompt-cache or platform response-cache hit, the cache-read portion is billed at cache_read_price, usually far below the input price. The X-TT-Cache-* headers show the savings.

Troubleshooting

Getting 401?

Wrong key, extra whitespace, or disabled. Check Authorization: Bearer sk-....

Getting 402?

Out of balance (billing_shortfall) or over the key's monthly limit (api_key_quota_exceeded). Recharge or adjust the limit.

Getting 429?

Rate limited. Retry after the Retry-After header, with exponential backoff.

How do I report an issue to support?

Bring the X-TT-Request-ID response header, approximate time, model, and the full error body. See Errors & rate limits.

Security

My API key leaked — what now?

Rotate the key in the console immediately; the old plaintext is invalidated at once. Use different keys with their own monthly limits per app.

I lost the key plaintext.

The plaintext is shown only once at creation. If lost, rotate a new one.