Create a key for this connection.
Open the Codimixa console and create a key for your OpenClaw setup. Keep it out of shared configuration files and use the secret mechanism supported by your installed OpenClaw version.
Add the connection settings.
OpenClaw registers custom model connections under models.providers. Use openai-responses for GPT agent models, anthropic-messages for native Claude requests, and openai-completions for Chat Completions models such as DeepSeek.
| Field | Set to |
|---|---|
| Connection name | codimixa |
| models.providers.codimixa.baseUrl | https://api.codimixa.com/v1 |
| models.providers.codimixa.api | openai-responses for GPT-6 Astra and GPT-6.1 Sol |
| models.providers.codimixa.apiKey | Your Codimixa key through the supported secret reference |
| Model entry | Exact model ID and limits from its Codimixa model page |
Start with one reply, then one tool task.
Select the configured model and send a small prompt. Next, run a task that exercises the tools you plan to use. Inspect the result and the account’s usage before adding recurring tasks.
Check the connection when a request fails.
For authentication failures, check that the secret resolves and the key has access. For model failures, check the model ID and API format. Stop repeated attempts while correcting the configuration.
Carry the actual rate into your budget.
Use the chosen model’s Codimixa rates when estimating your agent’s work. Selected model rates start at 0.3× official API prices; check the model and billing category.
A FEW DETAILS
Before you continue
Do I need a separate key from a model’s official API?
Requests to Codimixa use your Codimixa key. The key should match the service configured in the base URL.
Which API setting should I use?
For GPT agent models, use openai-responses. For native Claude tool use and cache controls, use anthropic-messages. For a Chat Completions connection such as DeepSeek, use openai-completions.