Connect Cherry Studio to CaMeL Hub (GPT, Claude, Gemini, DeepSeek)
Cherry Studio is a free, open-source desktop AI client for Windows, macOS, and Linux that lets you chat with multiple model providers, run RAG-backed knowledge bases, translate text, and compare answers from different models in one window. Point it at CaMeL Hub and a single API key unlocks GPT, Claude, Gemini, and DeepSeek models through the same OpenAI-compatible endpoint โ no separate provider accounts, no juggling multiple dashboards, and no re-entering a key every time you want to try a different model family.
-
Create a CaMeL Hub API key
Sign in to the CaMeL Hub console and open the API keys page. Create a new key (or reuse an existing one) and copy it โ Cherry Studio needs it in the next step. -
Install Cherry Studio
Download Cherry Studio for your OS from the official site and install it. Adding a custom model provider doesn't require any Cherry Studio account or sign-up. -
Add CaMeL Hub as a model provider
Open Settings โ Model Provider (่ฎพ็ฝฎ โ ๆจกๅๆๅก), click Add Provider, choose the OpenAI provider type, and name it something like "CaMeL Hub". Paste your API key into the field named API Key, then set the field named Base URL / API Host to the address below โ it's the OpenAI-compatible endpoint that serves every model on CaMeL Hub.https://api.camel-hub.com/v1 -
Add the models you want to use
In the same provider panel, click Add Model and paste the exact model ID from the CaMeL Hub console's model list (model IDs are case-sensitive). A few good starting points across providers:claude-sonnet-5 gpt-5.5 deepseek-v4-flash gemini-3.5-flash gpt-image-2 -
Verify the connection
Click the check / verify button next to the provider, or just send a message in a new chat. A green check or a normal reply confirms the key and Base URL are both correct. -
(Optional) add Claude as a native Anthropic provider
If you'd rather expose Claude-specific behavior through Cherry Studio's Anthropic provider type instead of the OpenAI-compatible one, add a second provider of type Anthropic and set its API Host to the bare domain below, with no /v1 suffix โ Cherry Studio appends /v1/messages itself, the same way it talks to Anthropic's own api.anthropic.com.https://api.camel-hub.com -
Turn on vision / tool-use per model if needed
Cherry Studio can't always auto-detect a custom model's capabilities. If a model supports images or function calling, open its entry under the provider and manually enable the matching capability toggle โ otherwise Cherry Studio may not send images along with your prompt even though the model can read them.
Frequently asked questions
The model list stays empty, or key verification keeps failing even though the key looks correct.
First double-check the Base URL has no trailing typo or extra slash, and that your API key hasn't expired. If it still fails, this is almost always a network-path issue rather than a key issue: corporate firewalls, some VPN clients, and certain DNS-over-HTTPS setups can intercept or mangle HTTPS requests to api.camel-hub.com. Try switching networks or temporarily disabling the VPN, then verify again.
The model doesn't seem to see the image I attached.
Enable the vision capability toggle for that specific model entry in Cherry Studio's provider settings (see step 7). For image-preview or other multimodal variants that can't return images while streaming, also make sure streaming output is turned off for that model โ check the model's own page on CaMeL Hub to confirm.
Calling a Claude model returns an AI_APICallError.
This usually means the provider type and Base URL don't match. If you added Claude under the OpenAI-compatible provider, use https://api.camel-hub.com/v1 unchanged. If you added it under Cherry Studio's native Anthropic provider, the API Host must be the bare domain https://api.camel-hub.com (no /v1) โ Cherry Studio adds /v1/messages automatically for that provider type.
Which models should I pick in Cherry Studio?
For everyday chat and coding, claude-sonnet-5 or gpt-5.5 are solid defaults. For fast, low-cost replies use gemini-3.5-flash or deepseek-v4-flash. For image generation inside Cherry Studio's drawing feature, use gpt-image-2.