Skip to main content
Now imagine giving your agent a 10-page instruction booklet. In Basic or Advanced mode, the agent reads the full booklet before every single reply — even if only 1 page matters right now. Multi-Prompt fixes this. It breaks the booklet into small, separate sections called capabilities — like sticky notes. The agent only picks up the sticky note that is relevant right now. The result is an agent that replies faster (less to read each time), stays focused on the current moment of the call, and works far better for long, multi-stage conversations. In short: Basic and Advanced are like reading the full notebook every time. Multi-Prompt is the right sticky note at the right moment. Based on the conversation, the agent automatically switches to the most relevant capability.

​
The two default capabilities

When you create a Multi-Prompt agent, two capabilities are already there by default.
  1. Base Prompt. This sets the agent’s identity and behaviour for the whole call — it stays active the entire time. It usually includes the agent role, company information, tone of voice, speaking style, rules and guidelines, and objection handling. Example: defining the agent as a friendly support rep who always speaks politely and professionally.
  2. Start Capability. This holds the first task the agent does when the call begins — greeting the customer, introducing itself, checking if it is a good time to talk, and deciding the next step. Example: the agent greets the customer, introduces itself, and asks if they are free to talk. If yes, it switches to the next capability, like lead qualification or appointment booking.

​
Adding more capabilities

From there, you can create as many additional capabilities as your call flow needs — such as Lead Qualification, Product Information, Appointment Booking, Objection Handling, Payment Collection, Customer Support, Feedback Collection, or Call Closing. The agent moves between them on its own as the call moves forward. Each additional capability comes with its own settings:
  • Capability Name. The name of the capability, so you know what it is for. Example: Appointment Booking.
  • Message Type While Switching. What the agent says when it moves into this capability. Static means it says a custom message you provide; None means it switches without saying anything.
  • System Prompt. The instructions for how the agent should behave while this capability is active.
  • Overwrite Existing Prompt (ON/OFF). ON means the agent ignores the Base Prompt and follows only this capability’s prompt. OFF means it follows both the Base Prompt and this capability’s prompt.
  • Stick Capability (ON/OFF). ON means once the agent enters this capability, it stays in it for the rest of the call. OFF means it can switch to other capabilities later if needed.
  • Use Filler (ON/OFF). ON means the agent can use filler phrases like “Let me check that for you” while it processes. OFF means it replies directly, with no fillers.
  • LLM Configuration. Controls which AI model runs this capability. Primary is the main model; Secondary is a backup used if the primary one is down.
  • Temperature. Controls how creative the agent’s replies are. Low (0.1–0.3) gives more consistent, predictable replies. High (0.7–1.0) gives more creative, varied replies.
  • Functions. Add actions the agent can perform in this capability — like booking appointments, ending calls, sending webhooks, or transferring calls.
  • Post Call Variables. The information to pull out of the call when this capability ends. Example: Customer Name, Appointment Date, Budget.
When to use: Multi-Prompt mode is best for long scripts, multi-step conversations, and complex call flows.

​
In short

All three build the same agent — they only differ in how you write and organise the instructions.