Creating and Deploying a Voice Agent
This article walks through creating an Alpharun voice agent, writing its instructions, and publishing it live so it can start handling calls. For a deeper look at any step, see the linked articles along the way.
Create the agent
Go to the Agents page and create a new agent. You provide two things:
A name for the agent.
A template to start from — Blank, After hours, Collecting documents, or Customer support. Templates pre-fill the instructions with a solid starting point you can edit.
That's all it takes to create an agent. Everything about how it behaves is defined next, in its instructions.
Write the instructions
Open the agent's Instructions tab. This is where you define how the agent behaves, organized into five sections: Agent name, Role and personality, Caller context, Conversation flow, and Behavior and guardrails.
Because the instructions are the single biggest driver of call quality, we cover what to put in each section in its own article: Writing Your Voice Agent's Instructions.
Add resources and outputs
On the Resources & outputs tab you can configure:
Tools — actions the agent can take during a call (see Using Tools with Your Voice Agent).
Knowledge base — reference documents the agent can pull answers from.
Outcome — a result classification applied to each call.
Summary template — the format of the post-call summary.
Data extraction fields — structured data pulled from each conversation (see Getting Voice Agent Call Data Out).
Test before you publish
Use the Testing tab to run test calls and confirm the agent follows your instructions and sounds the way you want.
Publish
Your edits autosave to a draft. Live callers only ever reach the published version, so you can iterate freely. When you're ready, click Publish to promote your draft to live. You can also Discard changes to revert a draft back to the live version.
Route calls to the agent
Each agent has an assigned phone number, shown on its Overview tab. Transfer calls to that number and the agent will handle them.
Note: Voice, model, and language come from your company's voice configuration rather than per-agent settings. On some setups these are managed centrally, so those controls may not appear in the agent editor.
