AGENT STUDIO

Make it your own.

Give your assistant a purpose, a voice and the knowledge it needs. Then put it to the test.

All agents ↗

Choose a starting point

A template replaces the fields below. Save your current draft first if you want to keep it.

What should your agent do?

Define its role and the outcome it should help people reach.

Agent namePurposeModel

Choose a chat model. This builder configures instructions; it does not deploy a model or connect external tools.

1 / 4

Saving is optional. Your current draft and reference text stay in this browser until you delete them.

BUILT FROM YOUR CONFIGURATION

Instruction preview

This is the system message your model receives. Changes appear here as you type.

Instruction-based assistantglm-5.2-fp8

Agent: Customer support

Purpose:
Turn a customer question and your policies into a clear, considerate response.

Instructions:
You help a support team draft customer replies. Use the customer message and policies supplied in this conversation. Identify the customer's concern, explain the applicable next step, and write a concise, considerate draft. Ask for missing policy details rather than inventing refund rules, promises, account information, or actions. You have no access to customer accounts, ticket systems, or external tools in this template. Do not say you have changed an account, issued a refund, or sent a message. Mark any action that requires a human decision.

Response style:
Be clear, practical and balanced. Match the user's language unless asked otherwise.

Capabilities:
You are a prompt-configured assistant. Do not claim to browse, access private systems, call external tools, send messages, or perform actions outside this conversation. Explain when you need information from the user.

Test this agent

This action temporarily stores the configuration and reference text in this tab for up to 30 minutes. No model request is sent until you send a message in chat. API usage may be billed.

Save to your account

A saved agent gets its own endpoint. Your application calls it with your API key; the instructions stay on the server.

…