Skip to main content
This guide walks you through the fastest way to get your first RapidCall agent live. Follow these eight steps in order. By the end, you will have a working agent ready for test calls.

​
Step 1 - Create your agent

Go to Agents and click Create Agent. Give the agent a clear, descriptive name such as Sarah - Appointment Setter or Tom - Recruitment Outreach. You will then choose the agent type:
  • Single Prompt - recommended for most use cases,
  • Conversation Flow - better when you need tightly controlled branching logic.
If you are unsure, start with Single Prompt. It is the simplest option and works well for most first builds.

​
Step 2 - Write the prompt

The prompt is the core instruction set behind your agent. It defines who the agent is, what it is trying to achieve, how it should respond, and when the call should end. If you are starting from scratch, gather the materials your team already has, such as:
  • company overview,
  • product or service information,
  • existing call scripts,
  • FAQs,
  • and objection-handling notes.
You can then use those materials to draft a first prompt. A strong prompt typically includes five sections:
  • Role
  • Context
  • Availability
  • Guidelines
  • Main Script
For a full breakdown of how to structure this, see the Agent Prompt section.

​
Step 3 - Select a voice stack

Next, choose how your agent will listen, think, and speak. RapidCall supports three main voice stack options:
  • Gemini Studio - a strong option for English-language agents with moderate cost and reliable performance,
  • OpenAI Realtime - best when very low latency is the top priority,
  • Custom Stack - best for more advanced use cases that need greater flexibility or more complex logic.
As a general rule:
  • use Gemini Studio for simpler English-language workflows,
  • use OpenAI Realtime for speed-sensitive use cases,
  • use Custom Stack for more complex production agents.
For deeper guidance, see the Voice Stack section.

​
Step 4 - Pick a voice

Open the Voice Gallery and listen to the available voice previews. Choose a voice that fits your use case, target audience, and region. Accent, tone, and delivery matter. A voice that works for a healthcare receptionist may not suit recruitment, collections, or real estate outreach. If you need to operate in multiple languages, it is usually better to create one agent per language rather than rely on a single multilingual setup.

​
Step 5 - Set up the knowledge base

Add the factual information your agent may need during calls. You can upload up to 10 files in PDF, DOC, or TXT format, with a maximum size of 100 MB per file, or paste content directly into the platform. Use the knowledge base for content such as:
  • product details,
  • pricing,
  • policies,
  • FAQs,
  • and opening hours.
Do not place behavioural instructions or scripts in the knowledge base. Those belong in the prompt. Keeping those two layers separate leads to better performance.

​
Step 6 - Configure call settings

Set the operational and conversational behaviour of the agent. This includes settings such as:
  • who speaks first,
  • first-message timing,
  • silence handling,
  • maximum call duration,
  • ring duration,
  • voicemail detection,
  • voice speed,
  • response eagerness,
  • interruption sensitivity,
  • backchannel behaviour,
  • and optional background sound.
These settings have a direct effect on how natural the agent feels in live calls, so they are worth tuning carefully.

​
Step 7 - Test the agent

Before connecting the agent to live phone traffic, test it thoroughly. Use the microphone or available test tools to hold real conversations with the agent. Try normal scenarios, but also try to break it. Ask unexpected questions, interrupt it, challenge the flow, and test edge cases. Most agents need several rounds of refinement before they are ready for production. That is normal. Each test call is logged, so you can review transcripts, recordings, and outcomes to improve the build.

​
Step 8 - Connect a phone number

Go to Phone Numbers and connect a number to the agent. You can either:
  • buy a new number directly through RapidCall,
  • or connect an existing business number through SIP trunking.
Once connected, assign the number to the agent for inbound and/or outbound use, configure routing, and set any relevant country permissions or caller identity options.

​
Need help?

The most technical parts of setup are usually telephony, CRM connectivity, and webhook configuration. If you want to move faster or avoid handling the first build alone, the RapidCall team can help scope, configure, and deploy your first agent with you.