TokenDance

Editor Review

If a project already uses an existing SDK, there is no need to write a separate client per model vendor. Change the base URL to the TokenDance gateway and keep sending the same chat-completion request. The console creates keys, including multiple keys and permissions, and the page says sign-in is one tap with Watcha. Billing is aggregated by token instead of topping up each vendor. If an endpoint fails, the same model can move to a backup provider, or one request can list fallback models. It fits application developers who want several models on one bill.

AI Tools Navigator Editorial TeamUpdated: 2026-10-09

What is TokenDance

TokenDance, also named in Chinese as a token gateway, describes itself as a unified model API gateway. Developers send requests to one gateway address and call through the OpenAI, Claude, or Gemini text protocols, with coverage for image, video, and text-to-speech as well. The system routes by model name to the matching provider, and bills token use in one account. One model can fail over across provider endpoints, and a single request can list fallback models. The page names providers including MiniMax, Qwen, Kimi, Zhipu, and DeepSeek

TokenDance Key Features

Several protocols, one gateway

One gateway address accepts OpenAI, Claude, and Gemini text protocols, and also covers image, video, and text-to-speech.

Route by model name

The model name in the request is sent to the matching provider, so callers do not keep a dispatch address per vendor.

Fail over across endpoints

One model can switch across provider endpoints, and one request can name several candidate models as fallbacks.

One token bill

Token use across providers is recorded on one bill, and the console can split permissions across multiple keys.

TokenDance Key Advantages

Changing the gateway address keeps the OpenAI, Claude, or Gemini text protocol, without rewriting the client

The provider is chosen from the model name, and image, video, and text-to-speech use the same entry

One model can move across provider endpoints, and one request can name fallback models

Token use is totaled across providers, and the console supports multiple keys and permissions

TokenDance Use Cases

Point an existing project at the gateway

A developer keeps the existing chat-completion code and only changes the gateway address and key to call models on the platform.

Call several providers on one bill

When an app needs Qwen, Kimi, Zhipu, or DeepSeek, token use is viewed in the same account.

Reroute when an endpoint fails

Configure a backup provider for the same model, or list candidate models in one request, so one dead endpoint does not fail the whole call.

Frequently Asked Questions

▶How much code has to change?

The documented setup replaces the gateway address and sends a key created in the console. Text requests can keep the OpenAI, Claude, or Gemini protocol, and image, video, and text-to-speech are on the same entry. You do not keep a separate address per provider. The page shows one-tap Watcha sign-in, and requests can be sent after the key exists.

▶What if one provider is down?

The page describes two backups. One model can use several provider endpoints and switch automatically, and one request can list candidate models that fail over in order. The bill still counts tokens in one place, rather than splitting accounts when the provider changes. Which endpoints exist depends on that model's console configuration at the time.

TokenDance

Editor Review

What is TokenDance

Basic Info

Best For

TokenDance Key Features

TokenDance Key Advantages

TokenDance Use Cases

Frequently Asked Questions

Latest News

User Reviews

Coze

SiliconFlow

码上飞

Zion

OpenRouter

秒哒

Dify

Trickle AI

TokenDance

Editor Review

What is TokenDance

Basic Info

Best For

TokenDance Key Features

TokenDance Key Advantages

TokenDance Use Cases

Frequently Asked Questions

Latest News

User Reviews

Related Tools

Coze

SiliconFlow

码上飞

Zion

OpenRouter

秒哒

Dify

Trickle AI