API overview
Base URL, authentication, supported endpoints, streaming and errors of the TokenFlux API
The TokenFlux API is an OpenAI-compatible gateway to many models. Your agents use the same gateway. One account and one balance pay for both.
Base URL and authentication
https://tokenflux.ai/v1Create a key on API keys and send it as a Bearer token:
Authorization: Bearer YOUR_TOKENFLUX_KEYMost SDKs need only the base URL and the key. Existing integrations can keep
the base URLs https://api.tokenflux.ai/openai/v1 and
https://api.tokenflux.ai/anthropic/v1; they reach the same endpoints.
Endpoints
| Endpoint | Compatible with |
|---|---|
GET /v1/models | OpenAI model list |
POST /v1/chat/completions | OpenAI Chat Completions |
POST /v1/responses | OpenAI Responses |
POST /v1/messages | Anthropic Messages |
POST /v1/embeddings | OpenAI Embeddings |
These are the documented endpoints. Do not depend on other paths.
Models
Use a model ID from GET /v1/models or from
Models & pricing. The models a key can call
depend on its access, so list them with the key you use.
Not every model supports every endpoint. A model's detail on Models & pricing
lists the protocols it supports. When a request reaches a model through a
different protocol, TokenFlux converts the request. Chat Completions and
Responses convert well. Other conversions can lose provider-specific features,
such as thinking signatures, cache_control or hosted tools. For those
features, call the endpoint that the model supports natively.
Streaming
Set stream to true to receive server-sent events in the format of the
compatible API. Closing the connection cancels the request.
Errors
Errors use the JSON error format of the compatible API with an HTTP error
status. If TokenFlux cannot reach the model service, it returns HTTP 502 with
this body:
{ "message": "Service temporarily unavailable" }Retry 429 and 5xx errors with backoff. Fix the request, the key or the
balance before you retry another 4xx error.
Billing
Each request is billed to the key's account at the prices on Models & pricing. Usage lists each request with its model, key, tokens and cost.