General steps
Almost every chat client speaks OpenAI-compatible protocols. Three steps:
- Pick OpenAI or OpenAI compatible as the provider.
- Set Base URL to
<your-gateway>/v1and API key to your OpenTokenRouter key. - Enter an available model ID (from
/v1/modelsor the Live settlement list on the home page).
Several clients support one-click import from the Download page (see each section below).
Cherry Studio
- Manual: Settings → Model service → Add → OpenAI compatible, fill in name, API address
<your-gateway>/v1, API key, and model ID. - One-click: the Cherry Studio import link on the Download page (
cherrystudio://protocol).
LobeChat
- Manual: Settings → AI service providers → OpenAI, fill in the API key, custom interface address
<your-gateway>/v1, and model ID. - One-click: the LobeChat example link on the Download page (with
?settings=params).
ChatBox
Settings → Models → provider OpenAI API, API domain <your-gateway>/v1, API key, then add a model ID.
NextChat (ChatGPT-Next-Web)
Settings → API address <your-gateway>/v1, API key, model (or default to use the gateway default).
Open WebUI
Admin → Settings → Connections → OpenAI API: Base URL <your-gateway>/v1, API key, save, then enable the models you need.
DeepChat
Use the DeepChat import link on the Download page (deepchat://provider/install), or add an OpenAI-compatible provider manually with Base URL <your-gateway>/v1.
AionUI
Use the AionUI import link on the Download page (aionui://provider/add).
OpenCat
Use the OpenCat join link on the Download page (opencat://team/join): domain <your-gateway>, token your key.
AMA
Use the AMA link on the Download page (ama://set-api-key) to set server and key in one click.
AI as Workspace
Use the AI as Workspace link on the Download page (aiaw.app/set-provider) to write an OpenAI-compatible provider in one click.
CC Switch and Fluent Read
- CC Switch: provider import on the Download page, or add an OpenAI-compatible provider in its panel.
- Fluent Read: use its built-in OpenAI-compatible configuration with
<your-gateway>/v1and your key.
FAQ
- Still cannot connect with /v1: make sure the URL is not double-appended (gateway root +
/v1) and has no extra path. - Empty model list: some clients require a manually entered model ID; copy one from
/v1/models. - 401: wrong key, or the key is disabled in the console.
- No streaming: disable response buffering in the client or a local proxy.