Glossary - Callers

Definitions of the terms you’ll see in the Callers dashboard, docs, and API.

A

Term	Definition
A2A connection	A link to an external AI agent over the Agent-to-Agent (A2A) protocol. Your Callers agent can pass the conversation to it and use its answers during a call. See A2A connections.
Action	A step that connects the agent to another system: an App Center action, a custom HTTP request, or custom code. Actions run before, during, or after a conversation. See Actions.
Agent	One AI worker with one job, such as confirming appointments or answering a support line. An agent holds its script, voice, channels, schedule, contacts, and results. Called “campaign” before September 2026, and still `campaign` in parts of the API. See Create an agent.
Agent group	A folder on the Agents page for organizing agents.
AI suggestions	Recommendations Callers generates from your agent’s results, such as steps with high drop-off or high interruption rates, or questions the agent missed.
App Center	The catalog of integrations (CRMs, calendars, spreadsheets, messaging, telephony) you connect once and use in actions and triggers. See App Center.
Availability	The hours and time zone in which an agent may call, plus its retry and concurrency rules. See Scheduling and retries.

B–C

Term	Definition
Background sound	Ambient audio, such as office noise, played under the agent’s voice on SIP calls. See Voices and sounds.
Block	A stage of the conversation on the script canvas, such as “Greeting” or “Qualify”. Also called a node. Blocks contain steps and can branch into subflows.
Call Data	Information about the current call, such as the contact’s phone number or the number the agent called from, that you can insert into steps and actions.
Channel	How an agent reaches people: Voice (phone and web calls), Text (SMS and WhatsApp), or Email.
Concurrency	How many conversations an agent runs at the same time. Outgoing voice agents default to 10 parallel calls (Max Outgoing Calls in Parallel).
Contact	A person the agent talks to, with a name, one or more phone numbers or an email address, and data for variables. See Uploading contacts.
Contact memory	Facts the AI remembers about a contact across conversations, used to personalize later calls. See Contact memory.
Contact status	Where a contact stands, such as New, Pending Retry, Interested, Not interested, Call later, or Do not call. See Statuses and outcomes.

D–I

Term	Definition
Dialing strategy	How an agent picks which of its numbers to call from: Rotate Numbers, Prefer Local Numbers, or Auto Optimize. See Phone numbers and caller ID.
Do Not Call (DNC) list	Numbers that must not be called. Callers keeps an account-level list and can check the US Federal DNC Registry and state lists. See Do Not Call.
DTMF / keypad input	Digits the contact types on their phone keypad, such as an account number. See Keypad input.
Draft	The editable version of an agent. Changes in a draft don’t affect live conversations until you publish.
Guardrails	Rules that check what the agent is about to say, such as “No medical advice”, and end the conversation, transfer to a human, or regenerate the reply when a rule is triggered. See Instructions and guardrails.
Human operator	A person on your team who takes transferred calls in the browser from a queue. See Human operators.
Insights Mode	A view of the script that shows, for each step, how many conversations reached it, hung up there, or interrupted. See Agent results and insights.
Instructions	Hidden guidance for the AI, set for the whole agent or inline in a step, that shapes behavior without being spoken.

K–P

Term	Definition
Key question	A structured question answered from the transcript after each conversation, such as “Budget” (number) or “Interested in a demo?” (yes/no). See Post-call analysis.
Knowledge base	Documents, web pages, and text the agent can search to answer questions. See Knowledge base.
Patience level	How long the agent waits after the contact stops talking before it replies, from Fast to Slow. See Conversation dynamics.
Post-call action	An action that runs after a conversation, often depending on its outcome, such as updating a CRM record.
Post-call analysis	The AI’s review of a finished conversation: a summary, answers to key questions, sentiment, and the success criteria result.
Pre-call chat	A text message sent before calling, asking whether the contact wants the call. See Pre-call chat.
Pre-conversation action	An action that runs before a call, before every attempt or only the first, usually to fetch data into variables.
Pronunciation dictionary	A list of words with how they should sound, used to fix how the agent says names and terms. See Pronunciation dictionary.
Publish	Make a saved draft the live version of an agent. See Versions and split traffic.

R–S

Term	Definition
Retry	Another attempt to reach a contact who didn’t answer, scheduled according to the agent’s retry rules.
Scenario	A response the agent can use at any point in the conversation when the contact says something that matches it, such as an objection or “call me later”. See Scenarios.
Script	The conversation flow of an agent, built from blocks and steps on a canvas. See Script editor.
SIP	The protocol used to connect your own carrier or PBX to Callers. See SIP integration.
Split traffic	Running several published versions of an agent at once with a share of conversations each, to compare results.
Step	A single thing the agent does inside a block: say or ask something, transfer the call, send an SMS or email, book a meeting, run an action, wait, or end the call. See Steps.
Studio Chat	The AI assistant in the dashboard that builds and edits agents and answers questions about your results. See Studio Chat.
Subaccount	A separate account under your main account, used by agencies for each client or by companies for each business unit. See Managing subaccounts.
Subflow	A branch of the script that the conversation follows depending on what the contact says.
Success criteria	The rules that decide whether a conversation reached its goal and how other outcomes are classified. Formerly “interest criteria”. See Success criteria.

T–W

Term	Definition
Test call	A call to your own phone or browser to try an agent before it goes live. See Testing your agent.
Transcript variable	A value the AI extracts from what the contact said, such as a date or a number, so it can be passed to an action.
Trigger	An event in a connected app that adds or archives a contact in an agent automatically. See Triggers.
Variable	A placeholder such as `{{name}}` in the script, filled for each contact from contact data, actions, or built-in values. See Variables.
Version	A saved state of an agent’s script and settings. Versions can be draft or published, and you can restore earlier ones.
Voicemail detection	Callers’ check, during and after the call, for whether a voicemail system answered. See Voicemail and screening.
Warm transfer	A transfer where the agent briefs the person receiving the call before connecting the contact. See Call transfers.
Webhook	An HTTP request Callers sends to your URL when a conversation event happens, such as a call finishing. See Webhooks.
White label	Presenting Callers to your clients under your own logo, name, and domain. See White label.
Widget	The embeddable component that lets website visitors talk or chat with your agent. See Installing the widget.

Core concepts

How agents, scripts, contacts, and results fit together.

FAQ

Short answers to common questions.

Quickstart

Build and test your first agent.

API introduction

Field names and conventions in the REST API.