Connect Cherry Studio

Use Cherry Studio’s custom provider settings to connect an existing desktop installation. Have your service API key ready and copy the service Base URL above.

1. Add a custom provider

Open Settings → Model Provider → + Add Provider. Give the provider a recognizable name, such as my-api, and enter your service key in API Key. Paste the key into the app privately, without a Bearer prefix.

In the current dialog, set the OpenAI endpoint to the service Base URL, including /v1. Check the request-path preview: it must end in /v1/chat/completions once. This guide uses OpenAI-compatible Chat Completions.

Older releases show an API Host field with automatic path handling. For that field, use the service Base URL ending in /v1/; the final slash tells Cherry Studio to append only chat/completions. Do not append a complete request path or # for this setup. See Cherry Studio’s URL rules.

Select Add and enable the provider using its top-right switch. Keep your other providers configured.

2. Add and select a model

Open the saved provider and select Sync models. Add the text models you want from the returned list. Fetching a list does not automatically add every model to the chat picker.

If you add one manually with +, use an exact ID returned by the authenticated models endpoint. A display name is not necessarily an API model ID.

In versions with a per-model Endpoint Type field, choose OpenAI for this Chat Completions setup. The separate OpenAI-Response and Anthropic choices use different request formats. Check the model’s settings after syncing if its detected protocol differs.

Return to chat, start a new conversation and choose the provider and model explicitly. Settings stay saved in the app; reopen the picker if an existing conversation retains its previous model.

3. Verify a text request

Send a short message and check the selected model and your service dashboard. Model Check also sends an inference request and can spend service balance.

For a missing model, check the provider switch and whether you added the synced model. For a 404, check the final URL. A successful chat does not verify images, embeddings, MCP tools or Cherry Agent, which has separate protocol requirements.

See the official custom provider guide and the reviewed model endpoint options.