Use ModelServer with Cursor
Cursor natively supports custom OpenAI-compatible endpoints, so pointing it at ModelServer is just filling in two fields — a base URL and an API key. The 3 steps below get you running, and then you can use Claude, GPT, Gemini, DeepSeek, and 40+ other models from Cursor's model picker.
Prerequisites
You'll need a ModelServer API key. If you don't have one yet, sign up and create one in the dashboard — the whole thing takes about 2 minutes.
1. Open Cursor's model settings
In Cursor, press Cmd/Ctrl + , to open Settings and select Models on the left. Scroll down to the OpenAI API Key section and expand Override OpenAI Base URL.
2. Fill in ModelServer's base URL and key
Paste the two values below into the corresponding inputs:
https://modelserver.dev/v1sk-ms-...3. Add custom models
In the Model Names section on the same page, click Add model and enter the model IDs you want to use. Here are some common ones — any model listed on ModelServer's pricing page can be entered as-is:
claude-opus-4 # Anthropic 旗舰
claude-sonnet-4 # 日常首选
claude-3-5-haiku # 快 & 便宜
gpt-5.2 # OpenAI 旗舰
gpt-4o # 速度 / 质量平衡
gemini-2.5-pro # 2M tokens 长上下文
deepseek-r1 # 开源推理
deepseek-v3 # 开源通用4. Verify
Back in the editor, press Cmd/Ctrl + K to open inline edit. Pick one of the models you just added from the dropdown (say, claude-sonnet-4) and type any prompt like "write fizzbuzz". If you get a reply, you're connected.
FAQ
Next steps
To learn how ModelServer handles tool calling under the hood, check Features → Tool Calling. Other editors (Cline, Continue, Roo Code) follow the same pattern — see the relevant guide under Integrations.