Set it up
1
Install Cline
Search for Cline in the VS Code extension marketplace and install it. A Cline icon appears in the activity bar.
2
Open settings
Click the gear icon at the top right of the Cline panel.
3
Enter the provider details
- API Provider:
OpenAI Compatible - Base URL:
https://api.laozhang.ai/v1 - API Key: your LaoZhang API key
- Model: the full model ID, such as
deepseek-v4-pro
OpenAI provider also accepts a custom base URL, but the settings and models on this page were checked with OpenAI Compatible, so use that one.4
Adjust model settings (optional)
In the model configuration, set Context Window and Max Output Tokens for the model you chose, and check Image Support if it accepts images. If you’re unsure, keep the defaults.
5
Test with a small task
Ask Cline to create a file and write one line to it. If it does, chat and tool calls both work.
Choose a model
Cline needs a model that returns tool calls reliably. These models have been confirmed to call tools in streaming Chat Completions requests:
Prices are in the model catalog.
FAQ
With GPT-6 I get “Function tools with reasoning_effort are not supported”
Before November 4, 2026, OpenAI’s rules let GPT-6 models call tools in Chat Completions only with reasoning effort set tonone, so Cline failed with this error. LaoZhang API now handles these requests: choose gpt-6.1-sol, gpt-6-sol, or another GPT-6 model and Cline can call tools with no extra settings.
If you still see the error, check that the base URL is https://api.laozhang.ai/v1, then send support the time of the error and the model.
I get “Unsupported parameter: ‘max_tokens’”
Before November 4, 2026, GPT-6 models rejectedmax_tokens in Chat Completions. LaoZhang API now accepts it, so you can keep Cline’s Max Output Tokens setting. If another model returns this error, clear Max Output Tokens and try again.
One task used a lot of tokens
Each step sends the conversation history, open files, and tool definitions. Give the token a quota and break large tasks into smaller ones. Models with a cache-read price, such asdeepseek-v4-pro, bill repeated context at the cache rate; see prompt cache billing.