CozeStudio

Use Cases

Configuration Steps

1. Open Model Management: On the device where Coze Studio is deployed, open http://localhost:8888/admin/#model-management . If Coze Studio is deployed on a server, replace localhost with the actual administration address.

2. Add a Model: Click “Add Model.”

3. Select the Protocol: Select openai as the Protocol.

4. Enter the Base URL: Enter https://api.maitoken.com/v1 . The model management page requires a Base URL, so you generally should not append /chat/completions.

5. Enter the API Key: Paste your MaiToken Key.

6. Enter the Model ID: Enter <Model ID> in the Model ID or Model field. After the model has been put into use, do not change its ID without careful consideration, as existing Agents may no longer be able to find it.

7. Configure Model Capabilities: Enable options such as Function Calling, Vision, and Reasoning according to the model's actual capabilities. If you are unsure, start by enabling text and tool-calling capabilities.

8. Test and Save: Click “Test.” After the test succeeds, save and enable the model.

9. Select the Model in an Agent: Create a test Agent, select the newly added MaiToken model from the model selector, save the configuration, and open the debugging page.

Test the Configuration

1. Test a Basic Conversation: Enter “Reply with OK only” in the chat box. If you receive a valid response, the API Key, API address, and Model ID are configured correctly.

2. Test Tool Calling: Enter “Call a simple tool. If no tools are currently available, reply with OK only.”

3. Test the Agent: Ask the Agent to perform a simple test task and confirm that the result is as expected. Do not use real production workloads for the initial test.

Common Issues