Connect an agent

Other OpenAI-compatible tools

Any tool supporting OpenAI-compatible or custom APIs can connect to Yomi API with three values: the API address, API key and model ID.

Prerequisites

API key: Create a key in the console's API Keys page, copy it and store it safely. It is visible only at creation.
Tool requirements: A custom connection setting, often labeled OpenAI Compatible, Custom API or Custom Provider.

Configure Yomi API

Enter the following three values in the tool's model or API settings:

Setting	Value	Description
API address	`https://api.yomiapi.com` or `https://api.yomiapi.com/v1`	See below for whether to include /v1
API Key	`YOUR_API_KEY`	A key created in API Keys
Model ID	`deepseek-v4-flash`, `claude-sonnet-5` or another model	Refer to the Model Catalog or `/v1/models` for current values

Whether to include /v1 in the API address

Tools construct request addresses differently. Check the example or placeholder in the address field:

If the example looks like https://api.openai.com/v1 (the full endpoint prefix is required), enter https://api.yomiapi.com/v1.
If the example looks like https://api.openai.com (the tool appends the /v1/... path itself), enter https://api.yomiapi.com.

In either case, the final complete request address should look like https://api.yomiapi.com/v1/chat/completions, with /v1 appearing exactly once.

Find an available model ID

Use a catalog ID for the model parameter (such as deepseek-v4-flash), not a display name such as GPT-5.6 mini. See Models (the Model Catalog) in the console for the complete list, or query the API directly:

curl https://api.yomiapi.com/v1/models \
  -H "Authorization: Bearer YOUR_API_KEY"

The API Keys page includes configuration examples for common tools. Use them alongside the three values in this article.

Start and verify

After configuration, use a minimal request to verify the key and address:

curl https://api.yomiapi.com/v1/chat/completions \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -d '{
    "model": "deepseek-v4-flash",
    "messages": [{"role": "user", "content": "Connection test: please reply ok"}]
  }'

A model reply in the returned JSON confirms that the connection works. Then start a conversation in the tool. A call record under Usage in the console confirms successful integration.

Common questions

What if the response is 404?

The final request address has a missing or duplicated /v1. Adjust it using the guidance under "Whether to include /v1 in the API address" and retry.

What if the response is 401?

The key is invalid or was not sent. Ensure it is the complete key created in API Keys, and that the tool sends it through the Authorization: Bearer request header.

What if the model does not exist?

The model name may be incorrect. Use a catalog ID, not a display name; refer to the Model Catalog or /v1/models response. Also confirm that your key's group includes the model.