Open the API settings in Cline.
Select OpenAI Compatible for a Chat Completions model. The example uses deepseek-v4.1-flash. For GPT-6.1 Sol agent tools, select a client mode that sends Responses requests.
| Cline setting | Value |
|---|---|
| API Provider | OpenAI Compatible |
| Base URL | https://api.codimixa.com/v1 |
| API Key | Your Codimixa API key |
| Model ID | deepseek-v4.1-flash |
Set model limits from the model page.
Use the selected model’s context limit, output limit and supported input capabilities. Where Cline asks for price inputs, use the current Codimixa rates and the units the field requires.
Try a small coding task.
Start with a read-only explanation of one function, then a scoped change that exercises a tool call. Review the change, run its test and compare the recorded account usage with your expectation.
Fix the field that caused the error.
Check key access for an authentication error, the exact model ID for a missing model, and the endpoint format for a connection error. If tool calls fail, check the model’s recorded capabilities.
Check the model rate before a longer session.
Selected model rates start at 0.3× official API pricing. A long task still needs a budget for its context, outputs and extra attempts.
A FEW DETAILS
Before you continue
Should I copy an official provider’s model ID?
Use the exact public model ID shown by Codimixa. A client must send the ID expected by the API service it is connected to.