Agent is typically translated as “intelligent agent.” Essentially, it is an AI that can autonomously execute tools. We believe that a good Agent has three key elements:

Model: The model is the Agent’s brain, determining the upper limit of the Agent’s capabilities.
Tools: Tools are the Agent’s hands and feet, determining the scope of the Agent’s capabilities.
Prompts: Prompts are the Agent’s instructions, determining the accuracy of the Agent’s capabilities.

At 302.AI, we have launched the Agent section. Simply configure the model + MCP + prompts to generate an Agent.

The generated Agent has 2 usage methods:

Can be used online through a chat-bot
Each Agent generates a unique model name that can be invoked in third-party clients through OpenAI-compatible Chat API

Online Testing

Click on the Agent name to launch a chatbot for online testing.

Note: This mode is for testing purposes only. It has no share code protection and no quota settings. If you need to share with other users, please refer to the next section.

Using in Chat-bot

The created Agent can be found in the model selection of the chatbot. Click on this model to create the corresponding chatbot.

In other chatbots, you can also use @ to find this Agent.

API Invocation

Click on the model name to view API invocation examples.

As you can see, it can be invoked through OpenAI-compatible Chat API. Simply change the model name to this Agent’s model name.

Note: This Agent cannot be invoked using API keys generated by other users.

In the response, the tool invocation process will be returned in the tool_call_content parameter. Developers can decide whether to display it to users.