Skip to main content
Before you start, check that your server meets the requirements.

​
Add an endpoint

Open Provider > Endpoints, select Add endpoint and fill in the form.
​
required
The catalog model your server runs.
​
required
The public HTTPS base URL of your OpenAI-compatible server, such as https://inference.example.com/v1. TokenChange adds /v1/chat/completions, /v1/completions and /v1/models itself, and a trailing /v1 is optional.The URL can’t contain credentials, a query string or a fragment, and its hostname must resolve only to public addresses.
​
Sent to your server as Authorization: Bearer .... Leave it empty if your server doesn’t need one. You can’t read the token back after saving it.
​
The model name your server expects in requests and lists in /v1/models. Defaults to the catalog slug.
​
required
Input, Output and Cached input, in USD per million tokens. See Pricing and earnings.
​
required
The most requests your server should handle at once. TokenChange never exceeds it, and endpoints with more slots receive a larger share of traffic.
​
The context length you serve, in tokens. Leave it empty to use the model’s default.
​
The weight format you serve, such as fp8. Customers can filter by it in the catalog.
​
The features your deployment supports: Tool calling, Structured output, Reasoning and Image input. Only select what your server actually supports.
Select Add endpoint to submit it for review. It shows as Awaiting approval until an operator reviews it. See Review and validation. You can also register endpoints with the REST API: POST /v1/provider/endpoints.

​
Manage an endpoint

Open an endpoint from Provider > Endpoints to see its connection details, validation results and earnings. Endpoints that are awaiting approval, rejected or suspended can’t be paused or resumed. Editing a rejected endpoint resubmits it for review. Price changes on an approved endpoint need operator approval. See Change your prices.