Skip to main content
You set up an agent in two ways. First, you give it a system prompt — its instruction manual. Second, you give it the pre-call variables it needs during the call, such as the user’s name, email, and address, based on what the call requires. The system prompt is the heart of the agent. It tells the agent who it is, what its goal is, how it should behave, what it must never do, the call flow to follow, and the objections it should handle. You can write this system prompt in three modes — Basic, Advanced, and Multi-Prompt. All three build the same thing: a smart AI calling agent. The only difference is how you write and organise the instructions. Let us go through them one by one.
If the agent connects to an external service, store API keys, tokens, and passwords in Settings → Secrets. Reference them as x_secrets.NAME; never place credential values directly in the prompt or agent configuration.